diff --git "a/log/log-train-2023-02-05-17-58-35-1" "b/log/log-train-2023-02-05-17-58-35-1" new file mode 100644--- /dev/null +++ "b/log/log-train-2023-02-05-17-58-35-1" @@ -0,0 +1,24984 @@ +2023-02-05 17:58:35,365 INFO [train.py:973] (1/4) Training started +2023-02-05 17:58:35,366 INFO [train.py:983] (1/4) Device: cuda:1 +2023-02-05 17:58:35,412 INFO [train.py:992] (1/4) {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'surt', 'icefall-git-sha1': 'b3d0d34-dirty', 'icefall-git-date': 'Sat Feb 4 14:53:48 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n07', 'IP address': '10.1.7.7'}, 'world_size': 4, 'master_port': 12354, 'tensorboard': True, 'num_epochs': 30, 'start_epoch': 1, 'start_batch': 0, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp/v1'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'base_lr': 0.05, 'lr_batches': 5000, 'lr_epochs': 3.5, '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': 10, 'average_period': 200, 'use_fp16': True, 'num_encoder_layers': '2,2,2,2,2', 'feedforward_dims': '768,768,768,768,768', 'nhead': '8,8,8,8,8', 'encoder_dims': '256,256,256,256,256', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '192,192,192,192,192', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'short_chunk_size': 50, 'num_left_chunks': 4, 'decode_chunk_len': 32, 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), '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, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'blank_id': 0, 'vocab_size': 500} +2023-02-05 17:58:35,412 INFO [train.py:994] (1/4) About to create model +2023-02-05 17:58:36,040 INFO [zipformer.py:402] (1/4) 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. +2023-02-05 17:58:36,052 INFO [train.py:998] (1/4) Number of model parameters: 20697573 +2023-02-05 17:58:51,148 INFO [train.py:1013] (1/4) Using DDP +2023-02-05 17:58:51,429 INFO [asr_datamodule.py:420] (1/4) About to get the shuffled train-clean-100, train-clean-360 and train-other-500 cuts +2023-02-05 17:58:52,644 INFO [asr_datamodule.py:224] (1/4) Enable MUSAN +2023-02-05 17:58:52,645 INFO [asr_datamodule.py:225] (1/4) About to get Musan cuts +2023-02-05 17:58:54,252 INFO [asr_datamodule.py:249] (1/4) Enable SpecAugment +2023-02-05 17:58:54,252 INFO [asr_datamodule.py:250] (1/4) Time warp factor: 80 +2023-02-05 17:58:54,253 INFO [asr_datamodule.py:260] (1/4) Num frame mask: 10 +2023-02-05 17:58:54,253 INFO [asr_datamodule.py:273] (1/4) About to create train dataset +2023-02-05 17:58:54,253 INFO [asr_datamodule.py:300] (1/4) Using DynamicBucketingSampler. +2023-02-05 17:58:54,274 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-05 17:58:56,355 INFO [asr_datamodule.py:316] (1/4) About to create train dataloader +2023-02-05 17:58:56,356 INFO [asr_datamodule.py:430] (1/4) About to get dev-clean cuts +2023-02-05 17:58:56,369 INFO [asr_datamodule.py:437] (1/4) About to get dev-other cuts +2023-02-05 17:58:56,393 INFO [asr_datamodule.py:347] (1/4) About to create dev dataset +2023-02-05 17:58:56,740 INFO [asr_datamodule.py:364] (1/4) About to create dev dataloader +2023-02-05 17:59:06,122 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-05 17:59:11,994 INFO [train.py:901] (1/4) Epoch 1, batch 0, loss[loss=7.264, simple_loss=6.579, pruned_loss=6.84, over 8251.00 frames. ], tot_loss[loss=7.264, simple_loss=6.579, pruned_loss=6.84, over 8251.00 frames. ], batch size: 24, lr: 2.50e-02, grad_scale: 2.0 +2023-02-05 17:59:11,994 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 17:59:24,182 INFO [train.py:935] (1/4) Epoch 1, validation: loss=6.888, simple_loss=6.229, pruned_loss=6.575, over 944034.00 frames. +2023-02-05 17:59:24,183 INFO [train.py:936] (1/4) Maximum memory allocated so far is 5591MB +2023-02-05 17:59:31,383 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=3.89 vs. limit=2.0 +2023-02-05 17:59:37,849 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-05 17:59:48,719 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=6.07 vs. limit=2.0 +2023-02-05 17:59:55,493 INFO [train.py:901] (1/4) Epoch 1, batch 50, loss[loss=1.38, simple_loss=1.223, pruned_loss=1.4, over 8329.00 frames. ], tot_loss[loss=2.187, simple_loss=1.977, pruned_loss=2.012, over 366493.11 frames. ], batch size: 25, lr: 2.75e-02, grad_scale: 0.25 +2023-02-05 17:59:56,152 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:00:12,635 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-05 18:00:13,733 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:00:28,695 INFO [train.py:901] (1/4) Epoch 1, batch 100, loss[loss=1.174, simple_loss=1.009, pruned_loss=1.308, over 8587.00 frames. ], tot_loss[loss=1.657, simple_loss=1.477, pruned_loss=1.628, over 642331.42 frames. ], batch size: 49, lr: 3.00e-02, grad_scale: 0.0625 +2023-02-05 18:00:28,822 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:00:32,052 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-05 18:00:32,810 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.087e+01 6.689e+01 1.862e+02 6.030e+02 6.185e+04, threshold=3.723e+02, percent-clipped=0.0 +2023-02-05 18:00:40,832 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=5.55 vs. limit=2.0 +2023-02-05 18:01:00,477 INFO [train.py:901] (1/4) Epoch 1, batch 150, loss[loss=1.087, simple_loss=0.9274, pruned_loss=1.156, over 8466.00 frames. ], tot_loss[loss=1.416, simple_loss=1.245, pruned_loss=1.443, over 862597.96 frames. ], batch size: 27, lr: 3.25e-02, grad_scale: 0.0625 +2023-02-05 18:01:20,192 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.76 vs. limit=2.0 +2023-02-05 18:01:34,585 INFO [train.py:901] (1/4) Epoch 1, batch 200, loss[loss=1.002, simple_loss=0.8486, pruned_loss=1.031, over 8363.00 frames. ], tot_loss[loss=1.277, simple_loss=1.112, pruned_loss=1.311, over 1032380.55 frames. ], batch size: 24, lr: 3.50e-02, grad_scale: 0.125 +2023-02-05 18:01:37,982 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.848e+01 5.119e+01 6.630e+01 8.708e+01 3.236e+02, threshold=1.326e+02, percent-clipped=1.0 +2023-02-05 18:02:05,442 INFO [train.py:901] (1/4) Epoch 1, batch 250, loss[loss=1.034, simple_loss=0.8716, pruned_loss=1.018, over 8584.00 frames. ], tot_loss[loss=1.188, simple_loss=1.027, pruned_loss=1.213, over 1159568.60 frames. ], batch size: 31, lr: 3.75e-02, grad_scale: 0.125 +2023-02-05 18:02:09,794 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=27.87 vs. limit=5.0 +2023-02-05 18:02:14,814 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-05 18:02:22,946 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-05 18:02:33,913 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=18.32 vs. limit=5.0 +2023-02-05 18:02:37,911 INFO [train.py:901] (1/4) Epoch 1, batch 300, loss[loss=0.8431, simple_loss=0.7, pruned_loss=0.8315, over 8085.00 frames. ], tot_loss[loss=1.129, simple_loss=0.968, pruned_loss=1.141, over 1264912.80 frames. ], batch size: 21, lr: 4.00e-02, grad_scale: 0.25 +2023-02-05 18:02:42,059 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=4.00 vs. limit=2.0 +2023-02-05 18:02:42,325 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=306.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:02:42,690 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 4.041e+01 5.570e+01 7.201e+01 9.677e+01 1.807e+02, threshold=1.440e+02, percent-clipped=6.0 +2023-02-05 18:02:46,519 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-02-05 18:02:47,401 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=314.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:03:00,231 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.35 vs. limit=2.0 +2023-02-05 18:03:10,248 INFO [train.py:901] (1/4) Epoch 1, batch 350, loss[loss=0.9444, simple_loss=0.7793, pruned_loss=0.9053, over 8308.00 frames. ], tot_loss[loss=1.09, simple_loss=0.9272, pruned_loss=1.089, over 1344291.99 frames. ], batch size: 25, lr: 4.25e-02, grad_scale: 0.25 +2023-02-05 18:03:26,549 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=5.76 vs. limit=5.0 +2023-02-05 18:03:42,311 INFO [train.py:901] (1/4) Epoch 1, batch 400, loss[loss=1.003, simple_loss=0.8237, pruned_loss=0.9328, over 8563.00 frames. ], tot_loss[loss=1.056, simple_loss=0.8917, pruned_loss=1.041, over 1400142.87 frames. ], batch size: 31, lr: 4.50e-02, grad_scale: 0.5 +2023-02-05 18:03:44,608 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=405.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:03:45,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 4.847e+01 5.714e+01 6.661e+01 8.261e+01 1.252e+02, threshold=1.332e+02, percent-clipped=0.0 +2023-02-05 18:03:55,269 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=421.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:04:05,708 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6695, 2.6772, 2.6447, 2.6648, 2.6531, 2.6787, 2.6743, 2.6711], + device='cuda:1'), covar=tensor([0.0066, 0.0065, 0.0061, 0.0071, 0.0066, 0.0063, 0.0065, 0.0070], + device='cuda:1'), in_proj_covar=tensor([0.0014, 0.0015, 0.0014, 0.0015, 0.0014, 0.0014, 0.0015, 0.0014], + device='cuda:1'), out_proj_covar=tensor([9.6141e-06, 9.7577e-06, 9.5819e-06, 9.3404e-06, 9.7828e-06, 9.3457e-06, + 9.8162e-06, 9.7360e-06], device='cuda:1') +2023-02-05 18:04:09,569 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8055, 3.8085, 3.8089, 3.7997, 3.7826, 3.8072, 3.8080, 3.8101], + device='cuda:1'), covar=tensor([0.0067, 0.0058, 0.0059, 0.0060, 0.0087, 0.0071, 0.0065, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0014, 0.0014, 0.0014, 0.0015, 0.0014, 0.0015, 0.0014, 0.0014], + device='cuda:1'), out_proj_covar=tensor([9.4092e-06, 9.2868e-06, 9.4074e-06, 9.0735e-06, 9.4270e-06, 9.4842e-06, + 9.4478e-06, 9.1931e-06], device='cuda:1') +2023-02-05 18:04:11,520 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=445.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:04:15,510 INFO [train.py:901] (1/4) Epoch 1, batch 450, loss[loss=0.9341, simple_loss=0.7604, pruned_loss=0.8563, over 7656.00 frames. ], tot_loss[loss=1.036, simple_loss=0.8687, pruned_loss=1.006, over 1450031.13 frames. ], batch size: 19, lr: 4.75e-02, grad_scale: 0.5 +2023-02-05 18:04:45,732 INFO [train.py:901] (1/4) Epoch 1, batch 500, loss[loss=1.02, simple_loss=0.8288, pruned_loss=0.905, over 8243.00 frames. ], tot_loss[loss=1.024, simple_loss=0.8527, pruned_loss=0.9766, over 1488734.62 frames. ], batch size: 24, lr: 4.99e-02, grad_scale: 1.0 +2023-02-05 18:04:49,468 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 5.283e+01 6.268e+01 7.626e+01 9.977e+01 2.238e+02, threshold=1.525e+02, percent-clipped=10.0 +2023-02-05 18:05:16,921 INFO [train.py:901] (1/4) Epoch 1, batch 550, loss[loss=0.9823, simple_loss=0.8068, pruned_loss=0.8238, over 8681.00 frames. ], tot_loss[loss=1.008, simple_loss=0.8357, pruned_loss=0.9405, over 1521153.63 frames. ], batch size: 34, lr: 4.98e-02, grad_scale: 1.0 +2023-02-05 18:05:22,218 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=560.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:05:33,859 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=580.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:05:39,262 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=586.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:05:47,865 INFO [train.py:901] (1/4) Epoch 1, batch 600, loss[loss=0.993, simple_loss=0.8195, pruned_loss=0.8007, over 8658.00 frames. ], tot_loss[loss=0.989, simple_loss=0.819, pruned_loss=0.8986, over 1540438.65 frames. ], batch size: 49, lr: 4.98e-02, grad_scale: 1.0 +2023-02-05 18:05:51,147 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 5.986e+01 8.101e+01 1.064e+02 1.512e+02 3.340e+02, threshold=2.128e+02, percent-clipped=22.0 +2023-02-05 18:05:51,950 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=608.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:05:57,484 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-05 18:06:12,877 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=8.27 vs. limit=5.0 +2023-02-05 18:06:15,537 INFO [train.py:901] (1/4) Epoch 1, batch 650, loss[loss=0.9258, simple_loss=0.7775, pruned_loss=0.6995, over 8492.00 frames. ], tot_loss[loss=0.9652, simple_loss=0.8002, pruned_loss=0.8506, over 1560934.51 frames. ], batch size: 39, lr: 4.98e-02, grad_scale: 1.0 +2023-02-05 18:06:20,642 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=658.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:06:31,074 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=677.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:06:44,418 INFO [train.py:901] (1/4) Epoch 1, batch 700, loss[loss=0.84, simple_loss=0.7033, pruned_loss=0.6257, over 8018.00 frames. ], tot_loss[loss=0.9383, simple_loss=0.7798, pruned_loss=0.8013, over 1573902.94 frames. ], batch size: 22, lr: 4.98e-02, grad_scale: 1.0 +2023-02-05 18:06:45,059 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=702.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:06:48,207 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 2.109e+02 3.132e+02 4.412e+02 1.990e+03, threshold=6.264e+02, percent-clipped=73.0 +2023-02-05 18:06:54,505 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-02-05 18:07:08,504 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3827, 0.7722, 1.1431, 1.8246, 0.9044, 1.2140, 1.8388, 1.4516], + device='cuda:1'), covar=tensor([0.7917, 1.8214, 1.3068, 0.4213, 1.1263, 0.7367, 0.5792, 0.7957], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0061, 0.0066, 0.0043, 0.0053, 0.0049, 0.0048, 0.0054], + device='cuda:1'), out_proj_covar=tensor([3.2304e-05, 4.5814e-05, 4.4696e-05, 2.2479e-05, 3.7465e-05, 2.7060e-05, + 2.7978e-05, 3.0013e-05], device='cuda:1') +2023-02-05 18:07:14,475 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=749.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:07:15,365 INFO [train.py:901] (1/4) Epoch 1, batch 750, loss[loss=0.8544, simple_loss=0.724, pruned_loss=0.6081, over 8470.00 frames. ], tot_loss[loss=0.9144, simple_loss=0.7622, pruned_loss=0.7568, over 1583359.21 frames. ], batch size: 25, lr: 4.97e-02, grad_scale: 1.0 +2023-02-05 18:07:25,616 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-05 18:07:26,842 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=773.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:07:32,308 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-05 18:07:43,635 INFO [train.py:901] (1/4) Epoch 1, batch 800, loss[loss=0.6687, simple_loss=0.5687, pruned_loss=0.4645, over 7652.00 frames. ], tot_loss[loss=0.8839, simple_loss=0.7399, pruned_loss=0.7089, over 1589716.37 frames. ], batch size: 19, lr: 4.97e-02, grad_scale: 2.0 +2023-02-05 18:07:46,599 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.447e+02 2.528e+02 3.354e+02 4.455e+02 1.086e+03, threshold=6.708e+02, percent-clipped=4.0 +2023-02-05 18:07:51,311 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=816.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:07:59,138 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-02-05 18:08:05,166 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=841.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:08:11,169 INFO [train.py:901] (1/4) Epoch 1, batch 850, loss[loss=0.8005, simple_loss=0.6832, pruned_loss=0.5439, over 8571.00 frames. ], tot_loss[loss=0.8585, simple_loss=0.7215, pruned_loss=0.6679, over 1597531.02 frames. ], batch size: 31, lr: 4.96e-02, grad_scale: 2.0 +2023-02-05 18:08:22,432 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=864.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:08:22,895 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=865.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:08:42,864 INFO [train.py:901] (1/4) Epoch 1, batch 900, loss[loss=0.6933, simple_loss=0.5956, pruned_loss=0.4587, over 7417.00 frames. ], tot_loss[loss=0.8313, simple_loss=0.7017, pruned_loss=0.6286, over 1597829.58 frames. ], batch size: 17, lr: 4.96e-02, grad_scale: 2.0 +2023-02-05 18:08:43,649 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=5.10 vs. limit=5.0 +2023-02-05 18:08:46,409 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 3.070e+02 3.818e+02 4.702e+02 7.623e+02, threshold=7.636e+02, percent-clipped=5.0 +2023-02-05 18:08:55,929 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=924.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:08:58,990 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=930.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:09:05,469 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=5.29 vs. limit=5.0 +2023-02-05 18:09:10,089 INFO [train.py:901] (1/4) Epoch 1, batch 950, loss[loss=0.7313, simple_loss=0.623, pruned_loss=0.4856, over 7800.00 frames. ], tot_loss[loss=0.8065, simple_loss=0.6835, pruned_loss=0.5934, over 1597289.85 frames. ], batch size: 20, lr: 4.96e-02, grad_scale: 2.0 +2023-02-05 18:09:10,748 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=952.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:09:26,428 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-05 18:09:37,674 INFO [train.py:901] (1/4) Epoch 1, batch 1000, loss[loss=0.7123, simple_loss=0.6225, pruned_loss=0.4461, over 8029.00 frames. ], tot_loss[loss=0.7851, simple_loss=0.6685, pruned_loss=0.5623, over 1600873.98 frames. ], batch size: 22, lr: 4.95e-02, grad_scale: 2.0 +2023-02-05 18:09:40,943 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.960e+02 3.215e+02 4.159e+02 4.799e+02 1.770e+03, threshold=8.319e+02, percent-clipped=6.0 +2023-02-05 18:09:52,916 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1029.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:09:53,907 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-05 18:09:59,182 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1039.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:10:02,590 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1045.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:10:05,090 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-05 18:10:05,577 INFO [train.py:901] (1/4) Epoch 1, batch 1050, loss[loss=0.7019, simple_loss=0.6116, pruned_loss=0.4381, over 8431.00 frames. ], tot_loss[loss=0.7677, simple_loss=0.657, pruned_loss=0.5355, over 1608888.76 frames. ], batch size: 49, lr: 4.95e-02, grad_scale: 2.0 +2023-02-05 18:10:07,174 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1054.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:10:14,045 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1067.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:10:33,038 INFO [train.py:901] (1/4) Epoch 1, batch 1100, loss[loss=0.737, simple_loss=0.6478, pruned_loss=0.4493, over 8361.00 frames. ], tot_loss[loss=0.7495, simple_loss=0.644, pruned_loss=0.5107, over 1612159.48 frames. ], batch size: 24, lr: 4.94e-02, grad_scale: 2.0 +2023-02-05 18:10:36,085 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 3.463e+02 4.480e+02 5.452e+02 1.232e+03, threshold=8.959e+02, percent-clipped=3.0 +2023-02-05 18:10:43,740 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1120.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:10:56,865 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1145.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:10:59,908 INFO [train.py:901] (1/4) Epoch 1, batch 1150, loss[loss=0.7116, simple_loss=0.6134, pruned_loss=0.4447, over 8511.00 frames. ], tot_loss[loss=0.7307, simple_loss=0.6305, pruned_loss=0.4875, over 1610057.59 frames. ], batch size: 28, lr: 4.94e-02, grad_scale: 2.0 +2023-02-05 18:11:01,587 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-05 18:11:11,750 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1171.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:11:27,719 INFO [train.py:901] (1/4) Epoch 1, batch 1200, loss[loss=0.673, simple_loss=0.5946, pruned_loss=0.4011, over 8499.00 frames. ], tot_loss[loss=0.7172, simple_loss=0.6213, pruned_loss=0.4687, over 1612006.46 frames. ], batch size: 28, lr: 4.93e-02, grad_scale: 4.0 +2023-02-05 18:11:30,964 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.986e+02 3.424e+02 4.173e+02 5.178e+02 8.029e+02, threshold=8.346e+02, percent-clipped=0.0 +2023-02-05 18:11:32,137 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1209.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:11:56,770 INFO [train.py:901] (1/4) Epoch 1, batch 1250, loss[loss=0.6584, simple_loss=0.5861, pruned_loss=0.3855, over 8495.00 frames. ], tot_loss[loss=0.703, simple_loss=0.6113, pruned_loss=0.451, over 1614366.47 frames. ], batch size: 26, lr: 4.92e-02, grad_scale: 4.0 +2023-02-05 18:12:21,158 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1295.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:12:24,253 INFO [train.py:901] (1/4) Epoch 1, batch 1300, loss[loss=0.7442, simple_loss=0.6555, pruned_loss=0.4404, over 8449.00 frames. ], tot_loss[loss=0.6911, simple_loss=0.6036, pruned_loss=0.4354, over 1619071.27 frames. ], batch size: 27, lr: 4.92e-02, grad_scale: 4.0 +2023-02-05 18:12:24,440 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1301.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:27,418 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 3.917e+02 4.747e+02 6.152e+02 9.080e+02, threshold=9.493e+02, percent-clipped=1.0 +2023-02-05 18:12:34,721 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1320.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:12:36,262 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1323.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:36,736 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1324.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:37,916 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1326.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:51,937 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1348.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:53,373 INFO [train.py:901] (1/4) Epoch 1, batch 1350, loss[loss=0.6178, simple_loss=0.5415, pruned_loss=0.3661, over 7822.00 frames. ], tot_loss[loss=0.6803, simple_loss=0.596, pruned_loss=0.4221, over 1618793.74 frames. ], batch size: 20, lr: 4.91e-02, grad_scale: 4.0 +2023-02-05 18:13:11,913 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 +2023-02-05 18:13:22,437 INFO [train.py:901] (1/4) Epoch 1, batch 1400, loss[loss=0.6174, simple_loss=0.5355, pruned_loss=0.3689, over 7789.00 frames. ], tot_loss[loss=0.6697, simple_loss=0.5888, pruned_loss=0.4095, over 1620891.95 frames. ], batch size: 19, lr: 4.91e-02, grad_scale: 4.0 +2023-02-05 18:13:25,821 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 3.466e+02 4.520e+02 5.912e+02 1.396e+03, threshold=9.040e+02, percent-clipped=6.0 +2023-02-05 18:13:50,929 INFO [train.py:901] (1/4) Epoch 1, batch 1450, loss[loss=0.6444, simple_loss=0.581, pruned_loss=0.3641, over 8496.00 frames. ], tot_loss[loss=0.6609, simple_loss=0.5829, pruned_loss=0.3989, over 1621344.63 frames. ], batch size: 28, lr: 4.90e-02, grad_scale: 4.0 +2023-02-05 18:13:51,608 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1452.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:13:54,962 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-05 18:14:21,306 INFO [train.py:901] (1/4) Epoch 1, batch 1500, loss[loss=0.6714, simple_loss=0.5894, pruned_loss=0.391, over 7142.00 frames. ], tot_loss[loss=0.6543, simple_loss=0.5787, pruned_loss=0.3903, over 1622200.61 frames. ], batch size: 72, lr: 4.89e-02, grad_scale: 4.0 +2023-02-05 18:14:24,736 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.084e+02 4.059e+02 4.884e+02 5.820e+02 1.191e+03, threshold=9.769e+02, percent-clipped=4.0 +2023-02-05 18:14:29,246 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1515.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:14:50,473 INFO [train.py:901] (1/4) Epoch 1, batch 1550, loss[loss=0.5674, simple_loss=0.5155, pruned_loss=0.3156, over 8086.00 frames. ], tot_loss[loss=0.6428, simple_loss=0.5709, pruned_loss=0.3788, over 1621015.10 frames. ], batch size: 21, lr: 4.89e-02, grad_scale: 4.0 +2023-02-05 18:15:08,627 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1580.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:15:10,873 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1584.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:15:18,490 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5878, 2.1465, 3.1350, 2.5066, 2.1341, 3.5923, 4.0077, 3.5393], + device='cuda:1'), covar=tensor([0.2579, 0.3282, 0.0491, 0.1811, 0.1837, 0.0334, 0.0178, 0.0334], + device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0140, 0.0077, 0.0117, 0.0104, 0.0069, 0.0062, 0.0081], + device='cuda:1'), out_proj_covar=tensor([9.1397e-05, 1.0646e-04, 4.3327e-05, 7.5971e-05, 7.0460e-05, 4.0048e-05, + 3.4240e-05, 4.6320e-05], device='cuda:1') +2023-02-05 18:15:20,755 INFO [train.py:901] (1/4) Epoch 1, batch 1600, loss[loss=0.6381, simple_loss=0.5732, pruned_loss=0.3587, over 8317.00 frames. ], tot_loss[loss=0.6364, simple_loss=0.5667, pruned_loss=0.3714, over 1623258.38 frames. ], batch size: 25, lr: 4.88e-02, grad_scale: 8.0 +2023-02-05 18:15:24,002 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1605.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:15:24,967 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.844e+02 4.893e+02 6.465e+02 8.597e+02 2.177e+03, threshold=1.293e+03, percent-clipped=12.0 +2023-02-05 18:15:31,426 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.6621, 1.3647, 3.4978, 2.4597, 3.5174, 3.1411, 3.0759, 3.3898], + device='cuda:1'), covar=tensor([0.0396, 0.3661, 0.0494, 0.1172, 0.0492, 0.0492, 0.0622, 0.0623], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0112, 0.0057, 0.0070, 0.0059, 0.0057, 0.0064, 0.0071], + device='cuda:1'), out_proj_covar=tensor([2.5403e-05, 7.3418e-05, 3.2264e-05, 4.6114e-05, 3.2357e-05, 3.3137e-05, + 3.7992e-05, 4.1670e-05], device='cuda:1') +2023-02-05 18:15:37,779 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1629.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:15:38,278 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1630.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:15:50,677 INFO [train.py:901] (1/4) Epoch 1, batch 1650, loss[loss=0.6742, simple_loss=0.6085, pruned_loss=0.3759, over 8602.00 frames. ], tot_loss[loss=0.6301, simple_loss=0.5627, pruned_loss=0.3643, over 1624583.62 frames. ], batch size: 39, lr: 4.87e-02, grad_scale: 8.0 +2023-02-05 18:16:21,963 INFO [train.py:901] (1/4) Epoch 1, batch 1700, loss[loss=0.5992, simple_loss=0.5303, pruned_loss=0.3402, over 8327.00 frames. ], tot_loss[loss=0.6191, simple_loss=0.5557, pruned_loss=0.3541, over 1620804.06 frames. ], batch size: 25, lr: 4.86e-02, grad_scale: 8.0 +2023-02-05 18:16:25,350 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.633e+02 4.287e+02 5.230e+02 6.455e+02 2.107e+03, threshold=1.046e+03, percent-clipped=2.0 +2023-02-05 18:16:51,248 INFO [train.py:901] (1/4) Epoch 1, batch 1750, loss[loss=0.6179, simple_loss=0.5631, pruned_loss=0.3393, over 8504.00 frames. ], tot_loss[loss=0.6135, simple_loss=0.5525, pruned_loss=0.348, over 1620375.83 frames. ], batch size: 28, lr: 4.86e-02, grad_scale: 8.0 +2023-02-05 18:17:06,014 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.35 vs. limit=2.0 +2023-02-05 18:17:18,058 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1796.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:17:21,123 INFO [train.py:901] (1/4) Epoch 1, batch 1800, loss[loss=0.5968, simple_loss=0.5495, pruned_loss=0.3237, over 8178.00 frames. ], tot_loss[loss=0.6056, simple_loss=0.5475, pruned_loss=0.3407, over 1615431.61 frames. ], batch size: 23, lr: 4.85e-02, grad_scale: 8.0 +2023-02-05 18:17:24,721 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.688e+02 4.554e+02 5.596e+02 6.733e+02 1.418e+03, threshold=1.119e+03, percent-clipped=4.0 +2023-02-05 18:17:52,108 INFO [train.py:901] (1/4) Epoch 1, batch 1850, loss[loss=0.4515, simple_loss=0.4298, pruned_loss=0.2364, over 7698.00 frames. ], tot_loss[loss=0.6001, simple_loss=0.5445, pruned_loss=0.335, over 1619605.13 frames. ], batch size: 18, lr: 4.84e-02, grad_scale: 8.0 +2023-02-05 18:17:55,072 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1856.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:18:06,748 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1875.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:18:13,282 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1886.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:18:14,316 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1888.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:18:21,902 INFO [train.py:901] (1/4) Epoch 1, batch 1900, loss[loss=0.5577, simple_loss=0.5209, pruned_loss=0.2976, over 8347.00 frames. ], tot_loss[loss=0.5974, simple_loss=0.5436, pruned_loss=0.3314, over 1619777.54 frames. ], batch size: 26, lr: 4.83e-02, grad_scale: 8.0 +2023-02-05 18:18:25,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.326e+02 4.483e+02 5.242e+02 7.443e+02 2.270e+03, threshold=1.048e+03, percent-clipped=7.0 +2023-02-05 18:18:27,946 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1911.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:18:27,957 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1911.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:18:37,733 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1928.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:18:45,003 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-05 18:18:52,608 INFO [train.py:901] (1/4) Epoch 1, batch 1950, loss[loss=0.5642, simple_loss=0.519, pruned_loss=0.305, over 8751.00 frames. ], tot_loss[loss=0.5912, simple_loss=0.5394, pruned_loss=0.3261, over 1619109.61 frames. ], batch size: 30, lr: 4.83e-02, grad_scale: 8.0 +2023-02-05 18:18:55,557 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-05 18:19:05,704 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1973.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:19:11,329 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-05 18:19:23,720 INFO [train.py:901] (1/4) Epoch 1, batch 2000, loss[loss=0.6076, simple_loss=0.5438, pruned_loss=0.3357, over 8487.00 frames. ], tot_loss[loss=0.5872, simple_loss=0.537, pruned_loss=0.3223, over 1620923.71 frames. ], batch size: 29, lr: 4.82e-02, grad_scale: 8.0 +2023-02-05 18:19:27,548 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.734e+02 4.600e+02 5.655e+02 7.771e+02 1.691e+03, threshold=1.131e+03, percent-clipped=5.0 +2023-02-05 18:19:50,354 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2043.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:19:56,661 INFO [train.py:901] (1/4) Epoch 1, batch 2050, loss[loss=0.5707, simple_loss=0.5353, pruned_loss=0.3031, over 8599.00 frames. ], tot_loss[loss=0.5766, simple_loss=0.5308, pruned_loss=0.3141, over 1620880.58 frames. ], batch size: 49, lr: 4.81e-02, grad_scale: 8.0 +2023-02-05 18:19:58,197 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6584, 1.8697, 3.4366, 2.2849, 2.3852, 4.6455, 4.4095, 3.9496], + device='cuda:1'), covar=tensor([0.3162, 0.4081, 0.0445, 0.2981, 0.1902, 0.0262, 0.0212, 0.0365], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0183, 0.0095, 0.0170, 0.0143, 0.0085, 0.0081, 0.0100], + device='cuda:1'), out_proj_covar=tensor([1.1921e-04, 1.3042e-04, 5.8222e-05, 1.1288e-04, 1.0164e-04, 5.3775e-05, + 4.7913e-05, 6.3234e-05], device='cuda:1') +2023-02-05 18:20:21,044 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2088.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:20:29,063 INFO [train.py:901] (1/4) Epoch 1, batch 2100, loss[loss=0.5327, simple_loss=0.5033, pruned_loss=0.2811, over 8105.00 frames. ], tot_loss[loss=0.5701, simple_loss=0.5269, pruned_loss=0.3088, over 1620693.73 frames. ], batch size: 21, lr: 4.80e-02, grad_scale: 16.0 +2023-02-05 18:20:32,707 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.532e+02 4.654e+02 5.875e+02 8.240e+02 2.515e+03, threshold=1.175e+03, percent-clipped=11.0 +2023-02-05 18:21:01,639 INFO [train.py:901] (1/4) Epoch 1, batch 2150, loss[loss=0.4822, simple_loss=0.4538, pruned_loss=0.2553, over 7788.00 frames. ], tot_loss[loss=0.5592, simple_loss=0.5207, pruned_loss=0.3005, over 1615957.76 frames. ], batch size: 19, lr: 4.79e-02, grad_scale: 16.0 +2023-02-05 18:21:09,732 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 +2023-02-05 18:21:11,748 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2167.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:21:29,899 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2192.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:21:35,015 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2200.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:21:35,570 INFO [train.py:901] (1/4) Epoch 1, batch 2200, loss[loss=0.6093, simple_loss=0.5555, pruned_loss=0.3316, over 8635.00 frames. ], tot_loss[loss=0.5532, simple_loss=0.518, pruned_loss=0.2955, over 1616363.20 frames. ], batch size: 39, lr: 4.78e-02, grad_scale: 16.0 +2023-02-05 18:21:39,329 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 3.885e+02 5.100e+02 6.280e+02 1.293e+03, threshold=1.020e+03, percent-clipped=3.0 +2023-02-05 18:21:46,988 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2219.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:21:55,779 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2232.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:22:07,881 INFO [train.py:901] (1/4) Epoch 1, batch 2250, loss[loss=0.5384, simple_loss=0.5328, pruned_loss=0.272, over 8255.00 frames. ], tot_loss[loss=0.5483, simple_loss=0.5156, pruned_loss=0.2915, over 1615790.84 frames. ], batch size: 24, lr: 4.77e-02, grad_scale: 16.0 +2023-02-05 18:22:30,951 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6680, 1.5983, 2.6430, 1.7471, 2.0147, 2.8844, 3.2924, 2.7620], + device='cuda:1'), covar=tensor([0.2397, 0.2919, 0.0431, 0.2504, 0.1359, 0.0375, 0.0254, 0.0407], + device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0183, 0.0098, 0.0169, 0.0146, 0.0084, 0.0081, 0.0099], + device='cuda:1'), out_proj_covar=tensor([1.2249e-04, 1.2969e-04, 6.5995e-05, 1.1360e-04, 1.0826e-04, 5.5481e-05, + 5.2542e-05, 6.3173e-05], device='cuda:1') +2023-02-05 18:22:41,043 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2299.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:22:42,094 INFO [train.py:901] (1/4) Epoch 1, batch 2300, loss[loss=0.5281, simple_loss=0.5111, pruned_loss=0.2726, over 8351.00 frames. ], tot_loss[loss=0.5453, simple_loss=0.5143, pruned_loss=0.2889, over 1620057.41 frames. ], batch size: 25, lr: 4.77e-02, grad_scale: 16.0 +2023-02-05 18:22:45,951 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.442e+02 5.272e+02 6.513e+02 7.975e+02 1.884e+03, threshold=1.303e+03, percent-clipped=9.0 +2023-02-05 18:22:51,189 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2315.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:22:56,956 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2324.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:23:03,193 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2334.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:23:08,004 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.41 vs. limit=5.0 +2023-02-05 18:23:09,679 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2344.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:23:12,276 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2347.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:23:14,702 INFO [train.py:901] (1/4) Epoch 1, batch 2350, loss[loss=0.5124, simple_loss=0.4917, pruned_loss=0.2666, over 8347.00 frames. ], tot_loss[loss=0.5412, simple_loss=0.5129, pruned_loss=0.2854, over 1623372.32 frames. ], batch size: 26, lr: 4.76e-02, grad_scale: 16.0 +2023-02-05 18:23:19,250 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2358.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:23:21,696 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7697, 1.5662, 1.6737, 2.1281, 1.3080, 1.3974, 1.1556, 2.0080], + device='cuda:1'), covar=tensor([0.2300, 0.2321, 0.1810, 0.0658, 0.2737, 0.2738, 0.3200, 0.2004], + device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0178, 0.0158, 0.0128, 0.0223, 0.0199, 0.0220, 0.0184], + device='cuda:1'), out_proj_covar=tensor([1.3616e-04, 1.3265e-04, 1.2580e-04, 8.9310e-05, 1.6231e-04, 1.4676e-04, + 1.6242e-04, 1.4031e-04], device='cuda:1') +2023-02-05 18:23:26,037 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2369.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:23:46,432 INFO [train.py:901] (1/4) Epoch 1, batch 2400, loss[loss=0.5055, simple_loss=0.5091, pruned_loss=0.2509, over 8495.00 frames. ], tot_loss[loss=0.5361, simple_loss=0.5097, pruned_loss=0.2817, over 1620268.61 frames. ], batch size: 29, lr: 4.75e-02, grad_scale: 16.0 +2023-02-05 18:23:50,350 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.591e+02 4.467e+02 5.905e+02 7.151e+02 1.301e+03, threshold=1.181e+03, percent-clipped=0.0 +2023-02-05 18:24:20,800 INFO [train.py:901] (1/4) Epoch 1, batch 2450, loss[loss=0.4867, simple_loss=0.4837, pruned_loss=0.2449, over 8027.00 frames. ], tot_loss[loss=0.5329, simple_loss=0.5077, pruned_loss=0.2794, over 1624811.77 frames. ], batch size: 22, lr: 4.74e-02, grad_scale: 16.0 +2023-02-05 18:24:52,776 INFO [train.py:901] (1/4) Epoch 1, batch 2500, loss[loss=0.4717, simple_loss=0.4476, pruned_loss=0.2478, over 7799.00 frames. ], tot_loss[loss=0.5288, simple_loss=0.5059, pruned_loss=0.2761, over 1623691.42 frames. ], batch size: 19, lr: 4.73e-02, grad_scale: 16.0 +2023-02-05 18:24:56,552 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.099e+02 5.238e+02 6.448e+02 8.237e+02 1.660e+03, threshold=1.290e+03, percent-clipped=6.0 +2023-02-05 18:24:57,984 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.9917, 1.0863, 3.9154, 1.8599, 3.5499, 3.2211, 3.3456, 3.2322], + device='cuda:1'), covar=tensor([0.0191, 0.3603, 0.0179, 0.1146, 0.0291, 0.0300, 0.0369, 0.0431], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0200, 0.0087, 0.0113, 0.0097, 0.0093, 0.0101, 0.0112], + device='cuda:1'), out_proj_covar=tensor([4.3750e-05, 1.2370e-04, 5.5997e-05, 7.6126e-05, 5.5256e-05, 5.2312e-05, + 6.1834e-05, 6.6717e-05], device='cuda:1') +2023-02-05 18:25:25,593 INFO [train.py:901] (1/4) Epoch 1, batch 2550, loss[loss=0.5002, simple_loss=0.4822, pruned_loss=0.2591, over 8205.00 frames. ], tot_loss[loss=0.526, simple_loss=0.5036, pruned_loss=0.2745, over 1619270.22 frames. ], batch size: 23, lr: 4.72e-02, grad_scale: 16.0 +2023-02-05 18:25:38,472 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2571.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:25:51,088 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2590.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:25:54,860 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2596.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:25:57,893 INFO [train.py:901] (1/4) Epoch 1, batch 2600, loss[loss=0.4493, simple_loss=0.469, pruned_loss=0.2148, over 8037.00 frames. ], tot_loss[loss=0.5219, simple_loss=0.5018, pruned_loss=0.2711, over 1618763.15 frames. ], batch size: 22, lr: 4.71e-02, grad_scale: 16.0 +2023-02-05 18:25:59,373 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2603.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:26:01,611 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.388e+02 4.352e+02 5.534e+02 7.344e+02 1.370e+03, threshold=1.107e+03, percent-clipped=3.0 +2023-02-05 18:26:06,880 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2615.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:26:15,230 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2628.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:26:25,262 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-02-05 18:26:31,162 INFO [train.py:901] (1/4) Epoch 1, batch 2650, loss[loss=0.4848, simple_loss=0.4899, pruned_loss=0.2398, over 8759.00 frames. ], tot_loss[loss=0.517, simple_loss=0.4992, pruned_loss=0.2676, over 1620654.11 frames. ], batch size: 30, lr: 4.70e-02, grad_scale: 16.0 +2023-02-05 18:26:38,477 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.09 vs. limit=2.0 +2023-02-05 18:27:03,808 INFO [train.py:901] (1/4) Epoch 1, batch 2700, loss[loss=0.4704, simple_loss=0.473, pruned_loss=0.2339, over 7979.00 frames. ], tot_loss[loss=0.512, simple_loss=0.4954, pruned_loss=0.2644, over 1617028.78 frames. ], batch size: 21, lr: 4.69e-02, grad_scale: 16.0 +2023-02-05 18:27:04,569 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2702.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:27:05,219 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2703.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:27:08,304 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.214e+02 4.351e+02 5.311e+02 6.408e+02 1.471e+03, threshold=1.062e+03, percent-clipped=4.0 +2023-02-05 18:27:28,815 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.5506, 3.7211, 3.2614, 1.3954, 3.1274, 3.4605, 3.2981, 3.0215], + device='cuda:1'), covar=tensor([0.0565, 0.0331, 0.0471, 0.2831, 0.0404, 0.0325, 0.0751, 0.0546], + device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0104, 0.0120, 0.0181, 0.0102, 0.0084, 0.0136, 0.0096], + device='cuda:1'), out_proj_covar=tensor([9.5898e-05, 8.7337e-05, 8.2219e-05, 1.2508e-04, 6.8701e-05, 6.0768e-05, + 1.0412e-04, 6.6109e-05], device='cuda:1') +2023-02-05 18:27:37,287 INFO [train.py:901] (1/4) Epoch 1, batch 2750, loss[loss=0.4997, simple_loss=0.4919, pruned_loss=0.2537, over 8292.00 frames. ], tot_loss[loss=0.5056, simple_loss=0.491, pruned_loss=0.2602, over 1609480.04 frames. ], batch size: 23, lr: 4.68e-02, grad_scale: 16.0 +2023-02-05 18:28:11,566 INFO [train.py:901] (1/4) Epoch 1, batch 2800, loss[loss=0.4342, simple_loss=0.4199, pruned_loss=0.2242, over 7685.00 frames. ], tot_loss[loss=0.5052, simple_loss=0.4917, pruned_loss=0.2595, over 1614745.18 frames. ], batch size: 18, lr: 4.67e-02, grad_scale: 16.0 +2023-02-05 18:28:15,258 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.340e+02 3.916e+02 4.898e+02 6.530e+02 2.276e+03, threshold=9.797e+02, percent-clipped=2.0 +2023-02-05 18:28:17,286 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4772, 1.8391, 2.7143, 1.6681, 2.1618, 3.1048, 3.3007, 2.6595], + device='cuda:1'), covar=tensor([0.2390, 0.1882, 0.0410, 0.2117, 0.1113, 0.0251, 0.0184, 0.0437], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0218, 0.0119, 0.0195, 0.0172, 0.0094, 0.0092, 0.0124], + device='cuda:1'), out_proj_covar=tensor([1.4755e-04, 1.5660e-04, 9.1568e-05, 1.3459e-04, 1.3457e-04, 6.8317e-05, + 6.8260e-05, 8.6717e-05], device='cuda:1') +2023-02-05 18:28:21,895 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2817.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:28:38,530 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2842.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:28:44,179 INFO [train.py:901] (1/4) Epoch 1, batch 2850, loss[loss=0.4473, simple_loss=0.4492, pruned_loss=0.2226, over 7654.00 frames. ], tot_loss[loss=0.5035, simple_loss=0.4905, pruned_loss=0.2583, over 1612584.52 frames. ], batch size: 19, lr: 4.66e-02, grad_scale: 16.0 +2023-02-05 18:29:18,794 INFO [train.py:901] (1/4) Epoch 1, batch 2900, loss[loss=0.6148, simple_loss=0.565, pruned_loss=0.3323, over 8606.00 frames. ], tot_loss[loss=0.4994, simple_loss=0.4878, pruned_loss=0.2556, over 1613116.18 frames. ], batch size: 49, lr: 4.65e-02, grad_scale: 16.0 +2023-02-05 18:29:22,681 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.417e+02 4.413e+02 5.664e+02 7.338e+02 1.737e+03, threshold=1.133e+03, percent-clipped=8.0 +2023-02-05 18:29:48,920 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-05 18:29:52,146 INFO [train.py:901] (1/4) Epoch 1, batch 2950, loss[loss=0.4434, simple_loss=0.4587, pruned_loss=0.2141, over 8333.00 frames. ], tot_loss[loss=0.4953, simple_loss=0.4856, pruned_loss=0.2526, over 1614109.44 frames. ], batch size: 25, lr: 4.64e-02, grad_scale: 16.0 +2023-02-05 18:29:54,921 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2955.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:30:25,914 INFO [train.py:901] (1/4) Epoch 1, batch 3000, loss[loss=0.4687, simple_loss=0.4641, pruned_loss=0.2367, over 7933.00 frames. ], tot_loss[loss=0.4924, simple_loss=0.4838, pruned_loss=0.2505, over 1612402.48 frames. ], batch size: 20, lr: 4.63e-02, grad_scale: 16.0 +2023-02-05 18:30:25,914 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 18:30:40,795 INFO [train.py:935] (1/4) Epoch 1, validation: loss=0.4518, simple_loss=0.5106, pruned_loss=0.1966, over 944034.00 frames. +2023-02-05 18:30:40,797 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6237MB +2023-02-05 18:30:43,954 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 +2023-02-05 18:30:44,896 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 4.264e+02 5.642e+02 7.781e+02 1.743e+03, threshold=1.128e+03, percent-clipped=6.0 +2023-02-05 18:31:07,309 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3037.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:31:08,851 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-02-05 18:31:13,909 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3047.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:31:14,751 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8149, 1.7292, 3.2529, 3.6106, 2.4020, 1.0424, 1.3845, 2.3946], + device='cuda:1'), covar=tensor([0.3407, 0.2444, 0.0284, 0.0232, 0.1132, 0.2511, 0.2182, 0.1349], + device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0148, 0.0076, 0.0094, 0.0158, 0.0162, 0.0155, 0.0170], + device='cuda:1'), out_proj_covar=tensor([1.2995e-04, 9.3687e-05, 4.7938e-05, 5.6883e-05, 9.7502e-05, 9.8520e-05, + 9.7091e-05, 1.0133e-04], device='cuda:1') +2023-02-05 18:31:16,515 INFO [train.py:901] (1/4) Epoch 1, batch 3050, loss[loss=0.4956, simple_loss=0.4842, pruned_loss=0.2535, over 7651.00 frames. ], tot_loss[loss=0.4896, simple_loss=0.482, pruned_loss=0.2486, over 1603401.73 frames. ], batch size: 19, lr: 4.62e-02, grad_scale: 16.0 +2023-02-05 18:31:20,856 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-02-05 18:31:30,842 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3073.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:31:47,475 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3098.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:31:49,293 INFO [train.py:901] (1/4) Epoch 1, batch 3100, loss[loss=0.4697, simple_loss=0.463, pruned_loss=0.2382, over 8146.00 frames. ], tot_loss[loss=0.4918, simple_loss=0.4836, pruned_loss=0.25, over 1606945.37 frames. ], batch size: 22, lr: 4.61e-02, grad_scale: 16.0 +2023-02-05 18:31:53,107 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.570e+02 4.257e+02 6.045e+02 8.311e+02 2.838e+03, threshold=1.209e+03, percent-clipped=13.0 +2023-02-05 18:32:24,770 INFO [train.py:901] (1/4) Epoch 1, batch 3150, loss[loss=0.4562, simple_loss=0.4657, pruned_loss=0.2234, over 8476.00 frames. ], tot_loss[loss=0.4913, simple_loss=0.4836, pruned_loss=0.2495, over 1606515.37 frames. ], batch size: 25, lr: 4.60e-02, grad_scale: 16.0 +2023-02-05 18:32:32,222 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3162.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:32:47,645 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3186.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:32:50,018 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-02-05 18:32:55,518 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.36 vs. limit=2.0 +2023-02-05 18:32:57,082 INFO [train.py:901] (1/4) Epoch 1, batch 3200, loss[loss=0.5802, simple_loss=0.5298, pruned_loss=0.3153, over 6696.00 frames. ], tot_loss[loss=0.491, simple_loss=0.4836, pruned_loss=0.2492, over 1605274.94 frames. ], batch size: 71, lr: 4.59e-02, grad_scale: 16.0 +2023-02-05 18:33:00,921 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.241e+02 4.232e+02 5.266e+02 6.948e+02 2.778e+03, threshold=1.053e+03, percent-clipped=2.0 +2023-02-05 18:33:07,793 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2011, 2.1950, 2.7909, 2.1002, 2.4877, 3.3838, 3.1102, 3.1037], + device='cuda:1'), covar=tensor([0.1702, 0.1992, 0.0432, 0.2018, 0.0942, 0.0200, 0.0238, 0.0275], + device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0234, 0.0126, 0.0215, 0.0171, 0.0097, 0.0096, 0.0124], + device='cuda:1'), out_proj_covar=tensor([1.5607e-04, 1.7023e-04, 9.9238e-05, 1.4916e-04, 1.3629e-04, 7.4761e-05, + 7.5616e-05, 9.1560e-05], device='cuda:1') +2023-02-05 18:33:32,110 INFO [train.py:901] (1/4) Epoch 1, batch 3250, loss[loss=0.4896, simple_loss=0.4933, pruned_loss=0.2429, over 8464.00 frames. ], tot_loss[loss=0.4925, simple_loss=0.4846, pruned_loss=0.2502, over 1606825.62 frames. ], batch size: 25, lr: 4.58e-02, grad_scale: 16.0 +2023-02-05 18:34:04,437 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3299.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:34:05,637 INFO [train.py:901] (1/4) Epoch 1, batch 3300, loss[loss=0.5469, simple_loss=0.5333, pruned_loss=0.2802, over 8703.00 frames. ], tot_loss[loss=0.49, simple_loss=0.4842, pruned_loss=0.2479, over 1616697.44 frames. ], batch size: 34, lr: 4.57e-02, grad_scale: 16.0 +2023-02-05 18:34:05,853 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3301.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:34:08,951 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3306.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:34:09,418 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 4.334e+02 5.638e+02 7.160e+02 2.697e+03, threshold=1.128e+03, percent-clipped=10.0 +2023-02-05 18:34:39,426 INFO [train.py:901] (1/4) Epoch 1, batch 3350, loss[loss=0.4214, simple_loss=0.4371, pruned_loss=0.2029, over 7251.00 frames. ], tot_loss[loss=0.4867, simple_loss=0.4824, pruned_loss=0.2456, over 1618118.10 frames. ], batch size: 16, lr: 4.56e-02, grad_scale: 16.0 +2023-02-05 18:35:01,937 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3381.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:35:14,988 INFO [train.py:901] (1/4) Epoch 1, batch 3400, loss[loss=0.4848, simple_loss=0.4841, pruned_loss=0.2427, over 8515.00 frames. ], tot_loss[loss=0.4845, simple_loss=0.4807, pruned_loss=0.2442, over 1617416.87 frames. ], batch size: 28, lr: 4.55e-02, grad_scale: 16.0 +2023-02-05 18:35:19,030 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.486e+02 3.960e+02 5.068e+02 6.311e+02 1.481e+03, threshold=1.014e+03, percent-clipped=3.0 +2023-02-05 18:35:23,813 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3414.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:35:26,543 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3418.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:35:43,715 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3443.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:35:48,748 INFO [train.py:901] (1/4) Epoch 1, batch 3450, loss[loss=0.4457, simple_loss=0.4406, pruned_loss=0.2254, over 7927.00 frames. ], tot_loss[loss=0.4845, simple_loss=0.48, pruned_loss=0.2445, over 1615127.64 frames. ], batch size: 20, lr: 4.54e-02, grad_scale: 16.0 +2023-02-05 18:36:12,075 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.41 vs. limit=5.0 +2023-02-05 18:36:21,044 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3496.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:36:24,211 INFO [train.py:901] (1/4) Epoch 1, batch 3500, loss[loss=0.4987, simple_loss=0.5081, pruned_loss=0.2447, over 8567.00 frames. ], tot_loss[loss=0.4835, simple_loss=0.4795, pruned_loss=0.2437, over 1611035.15 frames. ], batch size: 39, lr: 4.53e-02, grad_scale: 16.0 +2023-02-05 18:36:28,192 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.305e+02 4.405e+02 5.773e+02 7.537e+02 2.537e+03, threshold=1.155e+03, percent-clipped=7.0 +2023-02-05 18:36:36,262 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-05 18:36:48,246 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-02-05 18:36:57,801 INFO [train.py:901] (1/4) Epoch 1, batch 3550, loss[loss=0.5235, simple_loss=0.5051, pruned_loss=0.271, over 8470.00 frames. ], tot_loss[loss=0.4804, simple_loss=0.4782, pruned_loss=0.2413, over 1614206.40 frames. ], batch size: 25, lr: 4.51e-02, grad_scale: 16.0 +2023-02-05 18:37:02,065 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3557.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:37:07,160 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3564.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:37:19,199 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3582.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:37:33,290 INFO [train.py:901] (1/4) Epoch 1, batch 3600, loss[loss=0.5266, simple_loss=0.5114, pruned_loss=0.2709, over 8327.00 frames. ], tot_loss[loss=0.4849, simple_loss=0.481, pruned_loss=0.2444, over 1617559.46 frames. ], batch size: 25, lr: 4.50e-02, grad_scale: 16.0 +2023-02-05 18:37:37,965 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.853e+02 4.660e+02 6.337e+02 8.772e+02 4.832e+03, threshold=1.267e+03, percent-clipped=11.0 +2023-02-05 18:38:06,589 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3650.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:38:07,048 INFO [train.py:901] (1/4) Epoch 1, batch 3650, loss[loss=0.4941, simple_loss=0.4898, pruned_loss=0.2491, over 7810.00 frames. ], tot_loss[loss=0.4825, simple_loss=0.4795, pruned_loss=0.2428, over 1617290.08 frames. ], batch size: 20, lr: 4.49e-02, grad_scale: 16.0 +2023-02-05 18:38:18,964 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.4645, 1.2937, 1.0214, 0.2389, 0.9251, 0.9424, 0.1912, 1.2272], + device='cuda:1'), covar=tensor([0.0868, 0.0310, 0.0385, 0.1043, 0.0570, 0.0548, 0.1016, 0.0367], + device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0076, 0.0068, 0.0093, 0.0074, 0.0078, 0.0100, 0.0074], + device='cuda:1'), out_proj_covar=tensor([6.6076e-05, 5.0290e-05, 4.7154e-05, 7.0751e-05, 5.3792e-05, 5.4444e-05, + 7.0970e-05, 4.8331e-05], device='cuda:1') +2023-02-05 18:38:19,646 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3670.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:38:23,273 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 +2023-02-05 18:38:36,822 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3694.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:38:37,543 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3695.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:38:40,441 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-05 18:38:41,150 INFO [train.py:901] (1/4) Epoch 1, batch 3700, loss[loss=0.4382, simple_loss=0.4561, pruned_loss=0.2101, over 8479.00 frames. ], tot_loss[loss=0.4821, simple_loss=0.4797, pruned_loss=0.2423, over 1613096.62 frames. ], batch size: 25, lr: 4.48e-02, grad_scale: 16.0 +2023-02-05 18:38:45,146 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.178e+02 4.586e+02 6.278e+02 1.050e+03 3.437e+03, threshold=1.256e+03, percent-clipped=14.0 +2023-02-05 18:39:17,448 INFO [train.py:901] (1/4) Epoch 1, batch 3750, loss[loss=0.4096, simple_loss=0.4252, pruned_loss=0.197, over 7707.00 frames. ], tot_loss[loss=0.4792, simple_loss=0.4782, pruned_loss=0.2401, over 1617121.31 frames. ], batch size: 18, lr: 4.47e-02, grad_scale: 16.0 +2023-02-05 18:39:18,336 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3752.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:39:21,232 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.18 vs. limit=2.0 +2023-02-05 18:39:27,127 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3765.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:39:35,243 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3777.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:39:50,664 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.12 vs. limit=5.0 +2023-02-05 18:39:51,677 INFO [train.py:901] (1/4) Epoch 1, batch 3800, loss[loss=0.4199, simple_loss=0.4214, pruned_loss=0.2092, over 7428.00 frames. ], tot_loss[loss=0.4764, simple_loss=0.4762, pruned_loss=0.2383, over 1616008.21 frames. ], batch size: 17, lr: 4.46e-02, grad_scale: 16.0 +2023-02-05 18:39:55,874 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.457e+02 5.389e+02 6.979e+02 9.091e+02 1.609e+03, threshold=1.396e+03, percent-clipped=5.0 +2023-02-05 18:39:56,457 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 +2023-02-05 18:40:27,881 INFO [train.py:901] (1/4) Epoch 1, batch 3850, loss[loss=0.4614, simple_loss=0.4638, pruned_loss=0.2295, over 8076.00 frames. ], tot_loss[loss=0.4733, simple_loss=0.4736, pruned_loss=0.2365, over 1611899.57 frames. ], batch size: 21, lr: 4.45e-02, grad_scale: 16.0 +2023-02-05 18:40:46,565 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-05 18:41:01,009 INFO [train.py:901] (1/4) Epoch 1, batch 3900, loss[loss=0.5821, simple_loss=0.5339, pruned_loss=0.3152, over 7125.00 frames. ], tot_loss[loss=0.4719, simple_loss=0.4724, pruned_loss=0.2357, over 1612302.97 frames. ], batch size: 71, lr: 4.44e-02, grad_scale: 16.0 +2023-02-05 18:41:05,008 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.102e+02 5.552e+02 7.100e+02 9.321e+02 1.906e+03, threshold=1.420e+03, percent-clipped=2.0 +2023-02-05 18:41:05,748 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3908.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:41:29,958 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3944.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:41:35,351 INFO [train.py:901] (1/4) Epoch 1, batch 3950, loss[loss=0.4185, simple_loss=0.4419, pruned_loss=0.1975, over 8228.00 frames. ], tot_loss[loss=0.4708, simple_loss=0.4721, pruned_loss=0.2347, over 1613910.44 frames. ], batch size: 22, lr: 4.43e-02, grad_scale: 16.0 +2023-02-05 18:41:50,588 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-05 18:41:59,987 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9277, 1.1791, 3.1450, 1.3193, 2.4620, 3.4731, 3.3212, 3.2374], + device='cuda:1'), covar=tensor([0.1736, 0.2461, 0.0315, 0.2421, 0.0796, 0.0195, 0.0216, 0.0320], + device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0247, 0.0136, 0.0231, 0.0181, 0.0103, 0.0100, 0.0144], + device='cuda:1'), out_proj_covar=tensor([1.7220e-04, 1.8765e-04, 1.1712e-04, 1.6848e-04, 1.5655e-04, 8.3781e-05, + 8.8452e-05, 1.1598e-04], device='cuda:1') +2023-02-05 18:42:03,389 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3504, 1.6599, 1.2022, 1.4713, 1.4143, 1.5011, 1.1371, 1.8666], + device='cuda:1'), covar=tensor([0.1438, 0.0949, 0.2145, 0.0842, 0.1649, 0.1363, 0.2185, 0.0885], + device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0158, 0.0255, 0.0153, 0.0223, 0.0186, 0.0257, 0.0194], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 18:42:10,920 INFO [train.py:901] (1/4) Epoch 1, batch 4000, loss[loss=0.4952, simple_loss=0.5062, pruned_loss=0.2421, over 8474.00 frames. ], tot_loss[loss=0.4685, simple_loss=0.4709, pruned_loss=0.233, over 1610361.76 frames. ], batch size: 27, lr: 4.42e-02, grad_scale: 8.0 +2023-02-05 18:42:15,527 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.262e+02 4.572e+02 5.687e+02 7.371e+02 1.820e+03, threshold=1.137e+03, percent-clipped=4.0 +2023-02-05 18:42:24,567 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4021.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:42:25,844 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4023.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:42:36,384 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4038.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:42:42,707 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4046.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:42:46,086 INFO [train.py:901] (1/4) Epoch 1, batch 4050, loss[loss=0.4065, simple_loss=0.4421, pruned_loss=0.1854, over 8137.00 frames. ], tot_loss[loss=0.4687, simple_loss=0.471, pruned_loss=0.2332, over 1616932.60 frames. ], batch size: 22, lr: 4.41e-02, grad_scale: 8.0 +2023-02-05 18:43:22,339 INFO [train.py:901] (1/4) Epoch 1, batch 4100, loss[loss=0.4333, simple_loss=0.4452, pruned_loss=0.2106, over 7781.00 frames. ], tot_loss[loss=0.4675, simple_loss=0.4701, pruned_loss=0.2325, over 1619308.45 frames. ], batch size: 19, lr: 4.40e-02, grad_scale: 8.0 +2023-02-05 18:43:26,892 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.479e+02 4.889e+02 6.474e+02 8.616e+02 2.054e+03, threshold=1.295e+03, percent-clipped=5.0 +2023-02-05 18:43:56,551 INFO [train.py:901] (1/4) Epoch 1, batch 4150, loss[loss=0.383, simple_loss=0.413, pruned_loss=0.1765, over 8088.00 frames. ], tot_loss[loss=0.4634, simple_loss=0.4667, pruned_loss=0.2301, over 1614262.75 frames. ], batch size: 21, lr: 4.39e-02, grad_scale: 8.0 +2023-02-05 18:43:58,175 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4153.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:44:02,260 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4159.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:44:33,533 INFO [train.py:901] (1/4) Epoch 1, batch 4200, loss[loss=0.4116, simple_loss=0.4365, pruned_loss=0.1934, over 8094.00 frames. ], tot_loss[loss=0.4627, simple_loss=0.4666, pruned_loss=0.2294, over 1614387.72 frames. ], batch size: 21, lr: 4.38e-02, grad_scale: 8.0 +2023-02-05 18:44:38,305 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.396e+02 4.057e+02 5.109e+02 6.409e+02 1.525e+03, threshold=1.022e+03, percent-clipped=2.0 +2023-02-05 18:44:43,118 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2711, 1.2889, 2.8704, 1.1459, 2.0670, 3.2057, 3.1668, 2.8719], + device='cuda:1'), covar=tensor([0.2273, 0.2576, 0.0333, 0.2616, 0.0971, 0.0245, 0.0265, 0.0387], + device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0251, 0.0134, 0.0229, 0.0177, 0.0104, 0.0106, 0.0143], + device='cuda:1'), out_proj_covar=tensor([1.7406e-04, 1.9112e-04, 1.1864e-04, 1.6901e-04, 1.5586e-04, 8.5899e-05, + 9.4618e-05, 1.1781e-04], device='cuda:1') +2023-02-05 18:44:44,326 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-05 18:45:04,973 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-05 18:45:07,077 INFO [train.py:901] (1/4) Epoch 1, batch 4250, loss[loss=0.4317, simple_loss=0.431, pruned_loss=0.2162, over 7652.00 frames. ], tot_loss[loss=0.4607, simple_loss=0.4651, pruned_loss=0.2282, over 1610185.55 frames. ], batch size: 19, lr: 4.36e-02, grad_scale: 8.0 +2023-02-05 18:45:14,109 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.6373, 1.1663, 1.7408, 0.1668, 1.0392, 1.0210, 0.3513, 1.3622], + device='cuda:1'), covar=tensor([0.1055, 0.0613, 0.0272, 0.1602, 0.0772, 0.0648, 0.1298, 0.0403], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0086, 0.0072, 0.0110, 0.0080, 0.0097, 0.0106, 0.0081], + device='cuda:1'), out_proj_covar=tensor([7.8676e-05, 5.8146e-05, 5.1019e-05, 8.7177e-05, 6.2644e-05, 6.8205e-05, + 7.8033e-05, 5.5820e-05], device='cuda:1') +2023-02-05 18:45:26,612 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4279.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:45:29,997 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4253, 2.5060, 2.8087, 3.5377, 1.9861, 1.2789, 2.8148, 2.4395], + device='cuda:1'), covar=tensor([0.2250, 0.1223, 0.0587, 0.0314, 0.1198, 0.1657, 0.0911, 0.1270], + device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0099, 0.0072, 0.0078, 0.0102, 0.0117, 0.0124, 0.0114], + device='cuda:1'), out_proj_covar=tensor([9.0042e-05, 5.6350e-05, 3.9887e-05, 4.4295e-05, 5.7512e-05, 6.4370e-05, + 6.9393e-05, 6.1616e-05], device='cuda:1') +2023-02-05 18:45:33,325 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:45:33,449 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0367, 2.0750, 2.0407, 2.7772, 1.4009, 1.3419, 1.9020, 2.0699], + device='cuda:1'), covar=tensor([0.1188, 0.1371, 0.1153, 0.0268, 0.2164, 0.1994, 0.2003, 0.1432], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0271, 0.0248, 0.0165, 0.0324, 0.0302, 0.0345, 0.0268], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-02-05 18:45:42,875 INFO [train.py:901] (1/4) Epoch 1, batch 4300, loss[loss=0.3897, simple_loss=0.4101, pruned_loss=0.1847, over 7930.00 frames. ], tot_loss[loss=0.4583, simple_loss=0.4637, pruned_loss=0.2264, over 1610214.68 frames. ], batch size: 20, lr: 4.35e-02, grad_scale: 8.0 +2023-02-05 18:45:45,696 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4304.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:45:47,010 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4306.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:45:48,894 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.647e+02 4.666e+02 6.207e+02 8.078e+02 1.600e+03, threshold=1.241e+03, percent-clipped=6.0 +2023-02-05 18:46:18,302 INFO [train.py:901] (1/4) Epoch 1, batch 4350, loss[loss=0.4154, simple_loss=0.4138, pruned_loss=0.2085, over 7792.00 frames. ], tot_loss[loss=0.4533, simple_loss=0.4603, pruned_loss=0.2231, over 1611221.92 frames. ], batch size: 19, lr: 4.34e-02, grad_scale: 8.0 +2023-02-05 18:46:37,369 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-05 18:46:52,942 INFO [train.py:901] (1/4) Epoch 1, batch 4400, loss[loss=0.4444, simple_loss=0.4678, pruned_loss=0.2104, over 8346.00 frames. ], tot_loss[loss=0.4524, simple_loss=0.4593, pruned_loss=0.2227, over 1614902.56 frames. ], batch size: 26, lr: 4.33e-02, grad_scale: 8.0 +2023-02-05 18:46:54,542 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4403.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:46:57,934 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 4.338e+02 5.789e+02 7.262e+02 1.136e+03, threshold=1.158e+03, percent-clipped=0.0 +2023-02-05 18:46:58,965 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4409.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:47:18,621 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4434.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:47:21,210 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-05 18:47:29,977 INFO [train.py:901] (1/4) Epoch 1, batch 4450, loss[loss=0.4471, simple_loss=0.4561, pruned_loss=0.219, over 8360.00 frames. ], tot_loss[loss=0.4496, simple_loss=0.457, pruned_loss=0.2211, over 1607961.54 frames. ], batch size: 24, lr: 4.32e-02, grad_scale: 8.0 +2023-02-05 18:47:58,914 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.80 vs. limit=5.0 +2023-02-05 18:48:04,128 INFO [train.py:901] (1/4) Epoch 1, batch 4500, loss[loss=0.3921, simple_loss=0.4124, pruned_loss=0.1859, over 7791.00 frames. ], tot_loss[loss=0.4499, simple_loss=0.4564, pruned_loss=0.2216, over 1610012.42 frames. ], batch size: 19, lr: 4.31e-02, grad_scale: 8.0 +2023-02-05 18:48:05,597 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4503.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:48:09,058 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.251e+02 4.383e+02 5.863e+02 8.313e+02 2.632e+03, threshold=1.173e+03, percent-clipped=9.0 +2023-02-05 18:48:15,322 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-05 18:48:41,808 INFO [train.py:901] (1/4) Epoch 1, batch 4550, loss[loss=0.5309, simple_loss=0.5116, pruned_loss=0.2751, over 7132.00 frames. ], tot_loss[loss=0.4478, simple_loss=0.4554, pruned_loss=0.2201, over 1608685.53 frames. ], batch size: 71, lr: 4.30e-02, grad_scale: 8.0 +2023-02-05 18:49:02,396 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-05 18:49:16,692 INFO [train.py:901] (1/4) Epoch 1, batch 4600, loss[loss=0.4854, simple_loss=0.4943, pruned_loss=0.2383, over 8579.00 frames. ], tot_loss[loss=0.4498, simple_loss=0.457, pruned_loss=0.2213, over 1610645.05 frames. ], batch size: 31, lr: 4.29e-02, grad_scale: 8.0 +2023-02-05 18:49:21,478 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.209e+02 3.983e+02 5.037e+02 6.922e+02 1.236e+03, threshold=1.007e+03, percent-clipped=2.0 +2023-02-05 18:49:28,469 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4618.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:49:51,613 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4650.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:49:52,226 INFO [train.py:901] (1/4) Epoch 1, batch 4650, loss[loss=0.4273, simple_loss=0.4509, pruned_loss=0.2019, over 8024.00 frames. ], tot_loss[loss=0.4496, simple_loss=0.457, pruned_loss=0.221, over 1613104.70 frames. ], batch size: 22, lr: 4.28e-02, grad_scale: 8.0 +2023-02-05 18:49:59,118 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4659.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:50:16,191 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4684.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:50:27,583 INFO [train.py:901] (1/4) Epoch 1, batch 4700, loss[loss=0.534, simple_loss=0.5166, pruned_loss=0.2757, over 8335.00 frames. ], tot_loss[loss=0.4496, simple_loss=0.458, pruned_loss=0.2207, over 1615942.55 frames. ], batch size: 25, lr: 4.27e-02, grad_scale: 8.0 +2023-02-05 18:50:32,374 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.254e+02 4.576e+02 5.443e+02 6.674e+02 1.320e+03, threshold=1.089e+03, percent-clipped=4.0 +2023-02-05 18:51:01,880 INFO [train.py:901] (1/4) Epoch 1, batch 4750, loss[loss=0.3664, simple_loss=0.4132, pruned_loss=0.1598, over 8285.00 frames. ], tot_loss[loss=0.4455, simple_loss=0.4548, pruned_loss=0.2181, over 1614282.87 frames. ], batch size: 23, lr: 4.26e-02, grad_scale: 8.0 +2023-02-05 18:51:12,270 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4765.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:51:21,697 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-05 18:51:23,826 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-05 18:51:37,813 INFO [train.py:901] (1/4) Epoch 1, batch 4800, loss[loss=0.3653, simple_loss=0.3863, pruned_loss=0.1721, over 7691.00 frames. ], tot_loss[loss=0.4462, simple_loss=0.4555, pruned_loss=0.2185, over 1618583.38 frames. ], batch size: 18, lr: 4.25e-02, grad_scale: 8.0 +2023-02-05 18:51:42,622 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.690e+02 4.367e+02 5.327e+02 7.244e+02 1.939e+03, threshold=1.065e+03, percent-clipped=6.0 +2023-02-05 18:52:11,419 INFO [train.py:901] (1/4) Epoch 1, batch 4850, loss[loss=0.405, simple_loss=0.4378, pruned_loss=0.1862, over 8451.00 frames. ], tot_loss[loss=0.4485, simple_loss=0.4568, pruned_loss=0.2201, over 1620832.19 frames. ], batch size: 27, lr: 4.24e-02, grad_scale: 8.0 +2023-02-05 18:52:13,499 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-05 18:52:27,488 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4874.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:52:39,806 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4281, 1.4272, 1.9384, 1.5193, 1.2812, 2.0726, 0.5103, 1.0623], + device='cuda:1'), covar=tensor([0.0776, 0.0466, 0.0409, 0.0429, 0.0787, 0.0399, 0.1503, 0.0736], + device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0111, 0.0105, 0.0126, 0.0123, 0.0093, 0.0172, 0.0141], + device='cuda:1'), out_proj_covar=tensor([1.1046e-04, 9.3525e-05, 8.1902e-05, 9.4438e-05, 1.0104e-04, 7.2022e-05, + 1.3448e-04, 1.1395e-04], device='cuda:1') +2023-02-05 18:52:47,401 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4899.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:52:48,555 INFO [train.py:901] (1/4) Epoch 1, batch 4900, loss[loss=0.4193, simple_loss=0.4327, pruned_loss=0.2029, over 8096.00 frames. ], tot_loss[loss=0.446, simple_loss=0.455, pruned_loss=0.2186, over 1619512.72 frames. ], batch size: 23, lr: 4.23e-02, grad_scale: 8.0 +2023-02-05 18:52:53,376 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.332e+02 4.394e+02 5.447e+02 6.722e+02 1.310e+03, threshold=1.089e+03, percent-clipped=5.0 +2023-02-05 18:53:12,067 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7081, 1.9489, 2.4756, 1.7068, 1.5823, 2.4559, 0.9231, 1.3590], + device='cuda:1'), covar=tensor([0.0907, 0.0491, 0.0285, 0.0419, 0.0587, 0.0320, 0.1239, 0.0668], + device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0110, 0.0102, 0.0124, 0.0122, 0.0091, 0.0164, 0.0136], + device='cuda:1'), out_proj_covar=tensor([1.1115e-04, 9.2859e-05, 8.0234e-05, 9.3721e-05, 1.0132e-04, 7.0945e-05, + 1.2858e-04, 1.0974e-04], device='cuda:1') +2023-02-05 18:53:22,702 INFO [train.py:901] (1/4) Epoch 1, batch 4950, loss[loss=0.5099, simple_loss=0.4805, pruned_loss=0.2696, over 7926.00 frames. ], tot_loss[loss=0.4439, simple_loss=0.4537, pruned_loss=0.2171, over 1616002.06 frames. ], batch size: 20, lr: 4.21e-02, grad_scale: 8.0 +2023-02-05 18:53:59,112 INFO [train.py:901] (1/4) Epoch 1, batch 5000, loss[loss=0.4789, simple_loss=0.4671, pruned_loss=0.2454, over 8469.00 frames. ], tot_loss[loss=0.4444, simple_loss=0.4542, pruned_loss=0.2173, over 1618235.06 frames. ], batch size: 25, lr: 4.20e-02, grad_scale: 8.0 +2023-02-05 18:54:04,639 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.658e+02 4.358e+02 5.438e+02 7.182e+02 1.797e+03, threshold=1.088e+03, percent-clipped=3.0 +2023-02-05 18:54:12,968 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0484, 1.9157, 3.6931, 1.7154, 2.8544, 2.6865, 1.7062, 2.7190], + device='cuda:1'), covar=tensor([0.0866, 0.1574, 0.0158, 0.1111, 0.0862, 0.1237, 0.1168, 0.0970], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0236, 0.0192, 0.0250, 0.0286, 0.0304, 0.0247, 0.0275], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 18:54:13,642 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5021.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:54:30,644 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5046.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:54:33,891 INFO [train.py:901] (1/4) Epoch 1, batch 5050, loss[loss=0.3754, simple_loss=0.4257, pruned_loss=0.1625, over 7924.00 frames. ], tot_loss[loss=0.4432, simple_loss=0.4539, pruned_loss=0.2163, over 1618799.33 frames. ], batch size: 20, lr: 4.19e-02, grad_scale: 8.0 +2023-02-05 18:54:50,684 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-05 18:55:08,928 INFO [train.py:901] (1/4) Epoch 1, batch 5100, loss[loss=0.4606, simple_loss=0.4718, pruned_loss=0.2247, over 8098.00 frames. ], tot_loss[loss=0.4427, simple_loss=0.4528, pruned_loss=0.2163, over 1608852.81 frames. ], batch size: 23, lr: 4.18e-02, grad_scale: 8.0 +2023-02-05 18:55:13,603 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.507e+02 4.431e+02 5.257e+02 6.582e+02 1.311e+03, threshold=1.051e+03, percent-clipped=2.0 +2023-02-05 18:55:33,130 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.21 vs. limit=2.0 +2023-02-05 18:55:45,846 INFO [train.py:901] (1/4) Epoch 1, batch 5150, loss[loss=0.485, simple_loss=0.4943, pruned_loss=0.2378, over 8583.00 frames. ], tot_loss[loss=0.4411, simple_loss=0.4517, pruned_loss=0.2152, over 1611980.00 frames. ], batch size: 31, lr: 4.17e-02, grad_scale: 8.0 +2023-02-05 18:55:50,656 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5111, 1.4164, 2.2351, 0.5807, 2.0777, 1.9539, 0.6991, 1.9160], + device='cuda:1'), covar=tensor([0.0688, 0.0453, 0.0709, 0.1837, 0.1232, 0.0614, 0.2108, 0.0487], + device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0102, 0.0086, 0.0133, 0.0096, 0.0138, 0.0135, 0.0103], + device='cuda:1'), out_proj_covar=tensor([9.5556e-05, 7.3662e-05, 6.3745e-05, 1.0725e-04, 7.8101e-05, 1.0169e-04, + 1.0400e-04, 7.2532e-05], device='cuda:1') +2023-02-05 18:56:19,015 INFO [train.py:901] (1/4) Epoch 1, batch 5200, loss[loss=0.4194, simple_loss=0.4192, pruned_loss=0.2098, over 8286.00 frames. ], tot_loss[loss=0.4406, simple_loss=0.4514, pruned_loss=0.2149, over 1610845.15 frames. ], batch size: 23, lr: 4.16e-02, grad_scale: 8.0 +2023-02-05 18:56:23,576 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 3.937e+02 5.264e+02 6.479e+02 1.558e+03, threshold=1.053e+03, percent-clipped=7.0 +2023-02-05 18:56:51,636 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-05 18:56:55,099 INFO [train.py:901] (1/4) Epoch 1, batch 5250, loss[loss=0.4621, simple_loss=0.4748, pruned_loss=0.2247, over 8337.00 frames. ], tot_loss[loss=0.44, simple_loss=0.4514, pruned_loss=0.2143, over 1611727.52 frames. ], batch size: 25, lr: 4.15e-02, grad_scale: 8.0 +2023-02-05 18:57:28,844 INFO [train.py:901] (1/4) Epoch 1, batch 5300, loss[loss=0.5379, simple_loss=0.5225, pruned_loss=0.2767, over 7073.00 frames. ], tot_loss[loss=0.4403, simple_loss=0.4513, pruned_loss=0.2146, over 1609882.16 frames. ], batch size: 74, lr: 4.14e-02, grad_scale: 8.0 +2023-02-05 18:57:33,632 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.076e+02 4.278e+02 4.955e+02 6.641e+02 1.586e+03, threshold=9.909e+02, percent-clipped=4.0 +2023-02-05 18:57:33,850 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9093, 1.6606, 0.8840, 1.8833, 1.3636, 1.1374, 1.3096, 2.2293], + device='cuda:1'), covar=tensor([0.1118, 0.0734, 0.2474, 0.0567, 0.1791, 0.1367, 0.1596, 0.0506], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0188, 0.0308, 0.0222, 0.0280, 0.0233, 0.0299, 0.0247], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 18:57:51,140 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-05 18:58:04,337 INFO [train.py:901] (1/4) Epoch 1, batch 5350, loss[loss=0.4909, simple_loss=0.4757, pruned_loss=0.253, over 6710.00 frames. ], tot_loss[loss=0.4401, simple_loss=0.4512, pruned_loss=0.2145, over 1607093.55 frames. ], batch size: 71, lr: 4.13e-02, grad_scale: 8.0 +2023-02-05 18:58:39,824 INFO [train.py:901] (1/4) Epoch 1, batch 5400, loss[loss=0.4487, simple_loss=0.472, pruned_loss=0.2126, over 8323.00 frames. ], tot_loss[loss=0.4397, simple_loss=0.4511, pruned_loss=0.2141, over 1611672.27 frames. ], batch size: 25, lr: 4.12e-02, grad_scale: 8.0 +2023-02-05 18:58:44,298 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.977e+02 4.515e+02 5.788e+02 7.308e+02 1.362e+03, threshold=1.158e+03, percent-clipped=5.0 +2023-02-05 18:58:50,440 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.5030, 5.6810, 4.8305, 1.8117, 4.4811, 4.9441, 5.2012, 4.4580], + device='cuda:1'), covar=tensor([0.0653, 0.0256, 0.0674, 0.3425, 0.0406, 0.0460, 0.0646, 0.0458], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0174, 0.0198, 0.0262, 0.0159, 0.0129, 0.0197, 0.0120], + device='cuda:1'), out_proj_covar=tensor([1.8059e-04, 1.2781e-04, 1.3354e-04, 1.7272e-04, 1.0684e-04, 9.4440e-05, + 1.4481e-04, 8.5077e-05], device='cuda:1') +2023-02-05 18:59:13,402 INFO [train.py:901] (1/4) Epoch 1, batch 5450, loss[loss=0.4414, simple_loss=0.4528, pruned_loss=0.215, over 8532.00 frames. ], tot_loss[loss=0.4373, simple_loss=0.4499, pruned_loss=0.2124, over 1616049.11 frames. ], batch size: 28, lr: 4.11e-02, grad_scale: 8.0 +2023-02-05 18:59:41,788 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-05 18:59:49,950 INFO [train.py:901] (1/4) Epoch 1, batch 5500, loss[loss=0.4122, simple_loss=0.4331, pruned_loss=0.1957, over 7924.00 frames. ], tot_loss[loss=0.4354, simple_loss=0.4484, pruned_loss=0.2112, over 1613769.69 frames. ], batch size: 20, lr: 4.10e-02, grad_scale: 8.0 +2023-02-05 18:59:54,515 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.397e+02 4.451e+02 5.295e+02 6.340e+02 1.239e+03, threshold=1.059e+03, percent-clipped=2.0 +2023-02-05 19:00:23,639 INFO [train.py:901] (1/4) Epoch 1, batch 5550, loss[loss=0.3807, simple_loss=0.4081, pruned_loss=0.1766, over 8223.00 frames. ], tot_loss[loss=0.4365, simple_loss=0.4493, pruned_loss=0.2119, over 1613527.81 frames. ], batch size: 22, lr: 4.09e-02, grad_scale: 8.0 +2023-02-05 19:01:00,925 INFO [train.py:901] (1/4) Epoch 1, batch 5600, loss[loss=0.4324, simple_loss=0.4602, pruned_loss=0.2023, over 8256.00 frames. ], tot_loss[loss=0.4341, simple_loss=0.4478, pruned_loss=0.2102, over 1612820.50 frames. ], batch size: 24, lr: 4.08e-02, grad_scale: 8.0 +2023-02-05 19:01:05,769 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 3.916e+02 5.301e+02 6.582e+02 1.340e+03, threshold=1.060e+03, percent-clipped=3.0 +2023-02-05 19:01:34,544 INFO [train.py:901] (1/4) Epoch 1, batch 5650, loss[loss=0.4038, simple_loss=0.4253, pruned_loss=0.1911, over 7801.00 frames. ], tot_loss[loss=0.4338, simple_loss=0.4477, pruned_loss=0.2099, over 1616904.36 frames. ], batch size: 19, lr: 4.07e-02, grad_scale: 8.0 +2023-02-05 19:01:45,694 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-05 19:01:45,831 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=5668.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:02:09,329 INFO [train.py:901] (1/4) Epoch 1, batch 5700, loss[loss=0.3892, simple_loss=0.4029, pruned_loss=0.1877, over 7521.00 frames. ], tot_loss[loss=0.4366, simple_loss=0.4492, pruned_loss=0.212, over 1619911.50 frames. ], batch size: 18, lr: 4.06e-02, grad_scale: 8.0 +2023-02-05 19:02:15,263 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.140e+02 4.740e+02 5.744e+02 8.008e+02 1.790e+03, threshold=1.149e+03, percent-clipped=10.0 +2023-02-05 19:02:44,481 INFO [train.py:901] (1/4) Epoch 1, batch 5750, loss[loss=0.3654, simple_loss=0.388, pruned_loss=0.1714, over 7439.00 frames. ], tot_loss[loss=0.4349, simple_loss=0.4485, pruned_loss=0.2106, over 1618911.68 frames. ], batch size: 17, lr: 4.05e-02, grad_scale: 8.0 +2023-02-05 19:02:51,408 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-05 19:02:52,455 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.54 vs. limit=5.0 +2023-02-05 19:02:59,869 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=5773.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:03:19,628 INFO [train.py:901] (1/4) Epoch 1, batch 5800, loss[loss=0.5476, simple_loss=0.522, pruned_loss=0.2867, over 6742.00 frames. ], tot_loss[loss=0.4323, simple_loss=0.4465, pruned_loss=0.209, over 1614374.08 frames. ], batch size: 71, lr: 4.04e-02, grad_scale: 8.0 +2023-02-05 19:03:24,541 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.458e+02 3.671e+02 4.595e+02 5.667e+02 1.405e+03, threshold=9.190e+02, percent-clipped=2.0 +2023-02-05 19:03:57,256 INFO [train.py:901] (1/4) Epoch 1, batch 5850, loss[loss=0.4157, simple_loss=0.4478, pruned_loss=0.1918, over 8102.00 frames. ], tot_loss[loss=0.4298, simple_loss=0.4447, pruned_loss=0.2074, over 1614213.57 frames. ], batch size: 23, lr: 4.03e-02, grad_scale: 8.0 +2023-02-05 19:04:15,222 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=5876.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:04:32,490 INFO [train.py:901] (1/4) Epoch 1, batch 5900, loss[loss=0.4499, simple_loss=0.458, pruned_loss=0.2209, over 8590.00 frames. ], tot_loss[loss=0.4308, simple_loss=0.445, pruned_loss=0.2083, over 1614644.03 frames. ], batch size: 34, lr: 4.02e-02, grad_scale: 8.0 +2023-02-05 19:04:37,232 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.095e+02 4.155e+02 5.559e+02 6.668e+02 2.372e+03, threshold=1.112e+03, percent-clipped=6.0 +2023-02-05 19:05:03,786 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-02-05 19:05:09,347 INFO [train.py:901] (1/4) Epoch 1, batch 5950, loss[loss=0.3586, simple_loss=0.4002, pruned_loss=0.1585, over 8285.00 frames. ], tot_loss[loss=0.4282, simple_loss=0.4434, pruned_loss=0.2065, over 1617159.49 frames. ], batch size: 23, lr: 4.01e-02, grad_scale: 8.0 +2023-02-05 19:05:44,547 INFO [train.py:901] (1/4) Epoch 1, batch 6000, loss[loss=0.4452, simple_loss=0.4623, pruned_loss=0.214, over 8469.00 frames. ], tot_loss[loss=0.4288, simple_loss=0.4438, pruned_loss=0.2069, over 1614844.60 frames. ], batch size: 25, lr: 4.00e-02, grad_scale: 16.0 +2023-02-05 19:05:44,547 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 19:06:02,010 INFO [train.py:935] (1/4) Epoch 1, validation: loss=0.3351, simple_loss=0.4011, pruned_loss=0.1346, over 944034.00 frames. +2023-02-05 19:06:02,011 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6470MB +2023-02-05 19:06:06,794 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.323e+02 3.694e+02 4.999e+02 6.330e+02 1.596e+03, threshold=9.998e+02, percent-clipped=5.0 +2023-02-05 19:06:06,987 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6008.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:06:09,472 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6012.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:06:10,336 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=5.16 vs. limit=5.0 +2023-02-05 19:06:35,745 INFO [train.py:901] (1/4) Epoch 1, batch 6050, loss[loss=0.5053, simple_loss=0.5064, pruned_loss=0.2521, over 8332.00 frames. ], tot_loss[loss=0.4325, simple_loss=0.4459, pruned_loss=0.2095, over 1613765.38 frames. ], batch size: 25, lr: 3.99e-02, grad_scale: 8.0 +2023-02-05 19:06:36,629 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3090, 2.8783, 2.5633, 3.8699, 1.6172, 1.7346, 2.2029, 3.0009], + device='cuda:1'), covar=tensor([0.1369, 0.1484, 0.1215, 0.0202, 0.2417, 0.2089, 0.2150, 0.1198], + device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0313, 0.0291, 0.0183, 0.0348, 0.0333, 0.0383, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 19:06:42,877 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6061.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:06:45,814 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4324, 1.5649, 2.6830, 0.9943, 2.0183, 1.6059, 1.3358, 1.7839], + device='cuda:1'), covar=tensor([0.1352, 0.1357, 0.0291, 0.1675, 0.0950, 0.1917, 0.1352, 0.1062], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0265, 0.0236, 0.0290, 0.0335, 0.0341, 0.0284, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 19:07:05,560 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5081, 2.0194, 0.9701, 2.0427, 1.7079, 1.3838, 1.7481, 1.9613], + device='cuda:1'), covar=tensor([0.1403, 0.0704, 0.1911, 0.0824, 0.1126, 0.1329, 0.1556, 0.0896], + device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0211, 0.0327, 0.0242, 0.0299, 0.0260, 0.0318, 0.0257], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-02-05 19:07:10,269 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.8417, 0.9141, 0.9425, 0.9699, 0.6828, 0.9804, 0.0757, 0.5638], + device='cuda:1'), covar=tensor([0.0643, 0.0489, 0.0340, 0.0369, 0.0539, 0.0384, 0.1260, 0.0736], + device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0113, 0.0098, 0.0127, 0.0122, 0.0085, 0.0164, 0.0142], + device='cuda:1'), out_proj_covar=tensor([1.1441e-04, 1.0140e-04, 7.9299e-05, 1.0018e-04, 1.0693e-04, 6.8400e-05, + 1.3596e-04, 1.2118e-04], device='cuda:1') +2023-02-05 19:07:12,055 INFO [train.py:901] (1/4) Epoch 1, batch 6100, loss[loss=0.4615, simple_loss=0.4676, pruned_loss=0.2277, over 7974.00 frames. ], tot_loss[loss=0.43, simple_loss=0.4449, pruned_loss=0.2075, over 1615083.51 frames. ], batch size: 21, lr: 3.98e-02, grad_scale: 8.0 +2023-02-05 19:07:17,506 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.508e+02 4.942e+02 6.048e+02 7.564e+02 1.774e+03, threshold=1.210e+03, percent-clipped=15.0 +2023-02-05 19:07:23,142 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6117.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:07:28,993 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-05 19:07:29,787 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6127.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:07:44,056 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.3375, 5.6435, 4.7799, 1.8396, 4.7095, 5.1230, 4.8616, 4.1493], + device='cuda:1'), covar=tensor([0.0952, 0.0377, 0.0639, 0.4092, 0.0378, 0.0443, 0.1074, 0.0530], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0184, 0.0216, 0.0269, 0.0169, 0.0133, 0.0203, 0.0126], + device='cuda:1'), out_proj_covar=tensor([1.9402e-04, 1.3245e-04, 1.4140e-04, 1.7718e-04, 1.1049e-04, 9.4872e-05, + 1.4503e-04, 9.0425e-05], device='cuda:1') +2023-02-05 19:07:45,982 INFO [train.py:901] (1/4) Epoch 1, batch 6150, loss[loss=0.424, simple_loss=0.444, pruned_loss=0.2021, over 8535.00 frames. ], tot_loss[loss=0.4293, simple_loss=0.4441, pruned_loss=0.2073, over 1612551.82 frames. ], batch size: 31, lr: 3.97e-02, grad_scale: 8.0 +2023-02-05 19:07:47,401 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6153.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:07:58,547 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8032, 6.0211, 4.9435, 2.2793, 5.0120, 5.4277, 5.2681, 4.6362], + device='cuda:1'), covar=tensor([0.0607, 0.0298, 0.0706, 0.3144, 0.0320, 0.0411, 0.0894, 0.0390], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0181, 0.0213, 0.0265, 0.0166, 0.0132, 0.0200, 0.0123], + device='cuda:1'), out_proj_covar=tensor([1.8874e-04, 1.2984e-04, 1.3857e-04, 1.7487e-04, 1.0829e-04, 9.3786e-05, + 1.4285e-04, 8.8412e-05], device='cuda:1') +2023-02-05 19:08:12,168 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6188.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:22,942 INFO [train.py:901] (1/4) Epoch 1, batch 6200, loss[loss=0.4495, simple_loss=0.4576, pruned_loss=0.2207, over 8083.00 frames. ], tot_loss[loss=0.4296, simple_loss=0.4442, pruned_loss=0.2075, over 1612255.47 frames. ], batch size: 21, lr: 3.96e-02, grad_scale: 8.0 +2023-02-05 19:08:28,570 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.743e+02 4.155e+02 5.130e+02 7.106e+02 1.864e+03, threshold=1.026e+03, percent-clipped=2.0 +2023-02-05 19:08:36,391 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:37,176 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6221.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:42,651 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6229.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:44,638 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6232.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:57,431 INFO [train.py:901] (1/4) Epoch 1, batch 6250, loss[loss=0.5057, simple_loss=0.502, pruned_loss=0.2547, over 8483.00 frames. ], tot_loss[loss=0.4278, simple_loss=0.4435, pruned_loss=0.2061, over 1616582.80 frames. ], batch size: 28, lr: 3.95e-02, grad_scale: 8.0 +2023-02-05 19:09:20,021 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6284.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:09:32,703 INFO [train.py:901] (1/4) Epoch 1, batch 6300, loss[loss=0.4165, simple_loss=0.4489, pruned_loss=0.1921, over 8429.00 frames. ], tot_loss[loss=0.4263, simple_loss=0.4429, pruned_loss=0.2048, over 1618044.07 frames. ], batch size: 49, lr: 3.94e-02, grad_scale: 8.0 +2023-02-05 19:09:38,772 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.783e+02 4.352e+02 5.159e+02 6.362e+02 1.735e+03, threshold=1.032e+03, percent-clipped=4.0 +2023-02-05 19:09:56,807 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6335.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:10:07,347 INFO [train.py:901] (1/4) Epoch 1, batch 6350, loss[loss=0.389, simple_loss=0.4117, pruned_loss=0.1832, over 7697.00 frames. ], tot_loss[loss=0.4283, simple_loss=0.4446, pruned_loss=0.206, over 1619296.35 frames. ], batch size: 18, lr: 3.93e-02, grad_scale: 8.0 +2023-02-05 19:10:08,104 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6352.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:10:28,942 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6383.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:10:31,742 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-05 19:10:40,800 INFO [train.py:901] (1/4) Epoch 1, batch 6400, loss[loss=0.3563, simple_loss=0.3803, pruned_loss=0.1662, over 7542.00 frames. ], tot_loss[loss=0.4274, simple_loss=0.4443, pruned_loss=0.2052, over 1619432.42 frames. ], batch size: 18, lr: 3.92e-02, grad_scale: 8.0 +2023-02-05 19:10:43,626 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6405.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:10:45,789 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6408.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:10:46,251 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.785e+02 4.017e+02 4.991e+02 6.603e+02 1.156e+03, threshold=9.981e+02, percent-clipped=3.0 +2023-02-05 19:10:46,702 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-02-05 19:11:16,789 INFO [train.py:901] (1/4) Epoch 1, batch 6450, loss[loss=0.4947, simple_loss=0.4732, pruned_loss=0.2581, over 8249.00 frames. ], tot_loss[loss=0.4237, simple_loss=0.4414, pruned_loss=0.203, over 1620172.50 frames. ], batch size: 22, lr: 3.91e-02, grad_scale: 8.0 +2023-02-05 19:11:27,827 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:11:36,480 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6480.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:11:39,749 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6485.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:11:41,894 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6488.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:11:43,161 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5876, 1.7371, 1.7651, 2.6307, 1.0437, 1.2687, 1.5375, 1.7214], + device='cuda:1'), covar=tensor([0.1332, 0.1532, 0.1177, 0.0298, 0.2036, 0.1815, 0.2112, 0.1208], + device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0317, 0.0300, 0.0195, 0.0338, 0.0339, 0.0398, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 19:11:47,716 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6497.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:11:50,294 INFO [train.py:901] (1/4) Epoch 1, batch 6500, loss[loss=0.3586, simple_loss=0.3692, pruned_loss=0.174, over 6811.00 frames. ], tot_loss[loss=0.4225, simple_loss=0.4401, pruned_loss=0.2025, over 1612828.36 frames. ], batch size: 15, lr: 3.90e-02, grad_scale: 8.0 +2023-02-05 19:11:55,445 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.401e+02 4.204e+02 5.270e+02 6.161e+02 1.286e+03, threshold=1.054e+03, percent-clipped=6.0 +2023-02-05 19:11:58,505 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6513.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:03,293 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6520.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:11,141 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6532.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:14,878 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-05 19:12:25,065 INFO [train.py:901] (1/4) Epoch 1, batch 6550, loss[loss=0.3371, simple_loss=0.3916, pruned_loss=0.1413, over 8473.00 frames. ], tot_loss[loss=0.4227, simple_loss=0.4407, pruned_loss=0.2024, over 1616785.51 frames. ], batch size: 25, lr: 3.89e-02, grad_scale: 8.0 +2023-02-05 19:12:35,937 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6565.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:37,969 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-05 19:12:41,462 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6573.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:53,792 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6591.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:57,637 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-05 19:13:00,386 INFO [train.py:901] (1/4) Epoch 1, batch 6600, loss[loss=0.3886, simple_loss=0.4223, pruned_loss=0.1774, over 8469.00 frames. ], tot_loss[loss=0.4235, simple_loss=0.4414, pruned_loss=0.2029, over 1616998.24 frames. ], batch size: 27, lr: 3.89e-02, grad_scale: 8.0 +2023-02-05 19:13:05,674 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 4.035e+02 4.985e+02 6.404e+02 1.328e+03, threshold=9.970e+02, percent-clipped=3.0 +2023-02-05 19:13:07,910 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6612.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:10,619 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6616.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:18,640 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6628.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:31,559 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6647.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:31,597 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6647.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:34,156 INFO [train.py:901] (1/4) Epoch 1, batch 6650, loss[loss=0.4416, simple_loss=0.4617, pruned_loss=0.2107, over 8642.00 frames. ], tot_loss[loss=0.4202, simple_loss=0.439, pruned_loss=0.2007, over 1616858.13 frames. ], batch size: 49, lr: 3.88e-02, grad_scale: 8.0 +2023-02-05 19:13:36,498 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-02-05 19:13:53,673 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8571, 1.4908, 1.4643, 1.2162, 1.9388, 1.4125, 1.4964, 2.2620], + device='cuda:1'), covar=tensor([0.1760, 0.2509, 0.2856, 0.2718, 0.1305, 0.2338, 0.1741, 0.1153], + device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0299, 0.0299, 0.0291, 0.0280, 0.0269, 0.0272, 0.0262], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-02-05 19:13:56,264 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6680.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:01,298 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6688.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:09,924 INFO [train.py:901] (1/4) Epoch 1, batch 6700, loss[loss=0.4796, simple_loss=0.4655, pruned_loss=0.2468, over 7122.00 frames. ], tot_loss[loss=0.4216, simple_loss=0.4396, pruned_loss=0.2018, over 1612025.89 frames. ], batch size: 71, lr: 3.87e-02, grad_scale: 8.0 +2023-02-05 19:14:15,389 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.351e+02 4.140e+02 4.960e+02 6.260e+02 1.494e+03, threshold=9.921e+02, percent-clipped=3.0 +2023-02-05 19:14:25,019 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6723.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:38,630 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6743.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:42,143 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6748.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:43,045 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-02-05 19:14:44,012 INFO [train.py:901] (1/4) Epoch 1, batch 6750, loss[loss=0.3969, simple_loss=0.4205, pruned_loss=0.1867, over 8139.00 frames. ], tot_loss[loss=0.4206, simple_loss=0.4388, pruned_loss=0.2012, over 1611871.85 frames. ], batch size: 22, lr: 3.86e-02, grad_scale: 8.0 +2023-02-05 19:15:00,933 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6776.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:15:14,371 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-05 19:15:19,955 INFO [train.py:901] (1/4) Epoch 1, batch 6800, loss[loss=0.4679, simple_loss=0.481, pruned_loss=0.2274, over 8573.00 frames. ], tot_loss[loss=0.4202, simple_loss=0.4387, pruned_loss=0.2009, over 1613709.70 frames. ], batch size: 31, lr: 3.85e-02, grad_scale: 8.0 +2023-02-05 19:15:20,150 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6801.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:15:25,318 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 4.226e+02 5.434e+02 7.341e+02 1.725e+03, threshold=1.087e+03, percent-clipped=4.0 +2023-02-05 19:15:35,610 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6824.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:15:39,023 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6829.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:15:44,946 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-02-05 19:15:54,385 INFO [train.py:901] (1/4) Epoch 1, batch 6850, loss[loss=0.4166, simple_loss=0.4423, pruned_loss=0.1955, over 8364.00 frames. ], tot_loss[loss=0.4182, simple_loss=0.4373, pruned_loss=0.1996, over 1611828.39 frames. ], batch size: 24, lr: 3.84e-02, grad_scale: 8.0 +2023-02-05 19:16:04,831 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-05 19:16:06,402 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6868.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:23,437 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6893.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:25,426 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6896.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:16:29,307 INFO [train.py:901] (1/4) Epoch 1, batch 6900, loss[loss=0.4853, simple_loss=0.4828, pruned_loss=0.2439, over 8532.00 frames. ], tot_loss[loss=0.4198, simple_loss=0.4387, pruned_loss=0.2005, over 1616367.92 frames. ], batch size: 39, lr: 3.83e-02, grad_scale: 8.0 +2023-02-05 19:16:31,395 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6903.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:35,803 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.469e+02 3.796e+02 4.754e+02 6.076e+02 1.448e+03, threshold=9.507e+02, percent-clipped=2.0 +2023-02-05 19:16:48,742 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6927.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:49,392 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3640, 1.6924, 1.5287, 2.3754, 1.0454, 1.1045, 1.6067, 1.6785], + device='cuda:1'), covar=tensor([0.1500, 0.1707, 0.1572, 0.0454, 0.2182, 0.2539, 0.2067, 0.1221], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0321, 0.0303, 0.0200, 0.0336, 0.0331, 0.0395, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 19:16:49,409 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6928.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:54,880 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6936.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:56,903 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6939.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:00,425 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6944.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:00,474 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6944.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:05,075 INFO [train.py:901] (1/4) Epoch 1, batch 6950, loss[loss=0.3881, simple_loss=0.4176, pruned_loss=0.1794, over 8354.00 frames. ], tot_loss[loss=0.4185, simple_loss=0.4371, pruned_loss=0.1999, over 1617158.59 frames. ], batch size: 24, lr: 3.82e-02, grad_scale: 8.0 +2023-02-05 19:17:11,203 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-05 19:17:12,151 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6961.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:17,903 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6969.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:32,942 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6991.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:34,437 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4185, 1.6981, 1.1373, 2.0029, 1.5506, 1.1706, 1.0039, 1.9568], + device='cuda:1'), covar=tensor([0.1106, 0.0748, 0.1610, 0.0631, 0.1200, 0.1399, 0.1680, 0.0704], + device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0226, 0.0341, 0.0265, 0.0319, 0.0277, 0.0328, 0.0278], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-02-05 19:17:38,573 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6999.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:39,825 INFO [train.py:901] (1/4) Epoch 1, batch 7000, loss[loss=0.3953, simple_loss=0.4245, pruned_loss=0.1831, over 8593.00 frames. ], tot_loss[loss=0.4158, simple_loss=0.4353, pruned_loss=0.1982, over 1612886.64 frames. ], batch size: 31, lr: 3.81e-02, grad_scale: 8.0 +2023-02-05 19:17:45,245 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.380e+02 4.090e+02 4.918e+02 6.048e+02 1.151e+03, threshold=9.836e+02, percent-clipped=6.0 +2023-02-05 19:17:57,721 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7024.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:18:16,022 INFO [train.py:901] (1/4) Epoch 1, batch 7050, loss[loss=0.4042, simple_loss=0.4363, pruned_loss=0.186, over 8250.00 frames. ], tot_loss[loss=0.4164, simple_loss=0.436, pruned_loss=0.1984, over 1615156.83 frames. ], batch size: 22, lr: 3.80e-02, grad_scale: 8.0 +2023-02-05 19:18:26,088 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7066.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:18:50,235 INFO [train.py:901] (1/4) Epoch 1, batch 7100, loss[loss=0.3979, simple_loss=0.4218, pruned_loss=0.187, over 8751.00 frames. ], tot_loss[loss=0.4166, simple_loss=0.4363, pruned_loss=0.1984, over 1617646.91 frames. ], batch size: 30, lr: 3.79e-02, grad_scale: 8.0 +2023-02-05 19:18:53,871 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7106.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:18:55,757 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.243e+02 3.791e+02 4.613e+02 6.150e+02 1.722e+03, threshold=9.225e+02, percent-clipped=5.0 +2023-02-05 19:19:25,902 INFO [train.py:901] (1/4) Epoch 1, batch 7150, loss[loss=0.4873, simple_loss=0.458, pruned_loss=0.2583, over 7553.00 frames. ], tot_loss[loss=0.4157, simple_loss=0.4358, pruned_loss=0.1978, over 1618951.08 frames. ], batch size: 18, lr: 3.78e-02, grad_scale: 8.0 +2023-02-05 19:19:46,605 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7181.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:19:48,142 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-05 19:19:56,066 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7195.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:19:59,479 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7200.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:19:59,943 INFO [train.py:901] (1/4) Epoch 1, batch 7200, loss[loss=0.3514, simple_loss=0.372, pruned_loss=0.1654, over 7439.00 frames. ], tot_loss[loss=0.416, simple_loss=0.436, pruned_loss=0.198, over 1620531.18 frames. ], batch size: 17, lr: 3.78e-02, grad_scale: 8.0 +2023-02-05 19:20:05,322 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.325e+02 4.231e+02 5.262e+02 7.053e+02 1.685e+03, threshold=1.052e+03, percent-clipped=7.0 +2023-02-05 19:20:13,040 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:20:16,296 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7225.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:20:25,922 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7240.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:20:33,000 INFO [train.py:901] (1/4) Epoch 1, batch 7250, loss[loss=0.3839, simple_loss=0.4141, pruned_loss=0.1769, over 8347.00 frames. ], tot_loss[loss=0.4158, simple_loss=0.4366, pruned_loss=0.1975, over 1620269.74 frames. ], batch size: 24, lr: 3.77e-02, grad_scale: 8.0 +2023-02-05 19:20:48,481 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7271.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:21:03,382 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1430, 1.7615, 1.7355, 0.3547, 1.6512, 1.3204, 0.2150, 1.7288], + device='cuda:1'), covar=tensor([0.0397, 0.0240, 0.0235, 0.0817, 0.0306, 0.0533, 0.0710, 0.0206], + device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0117, 0.0098, 0.0158, 0.0111, 0.0177, 0.0162, 0.0128], + device='cuda:1'), out_proj_covar=tensor([1.1261e-04, 8.5500e-05, 7.7439e-05, 1.2152e-04, 9.1249e-05, 1.4085e-04, + 1.2374e-04, 9.6372e-05], device='cuda:1') +2023-02-05 19:21:03,978 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7293.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:21:08,993 INFO [train.py:901] (1/4) Epoch 1, batch 7300, loss[loss=0.431, simple_loss=0.4559, pruned_loss=0.203, over 8619.00 frames. ], tot_loss[loss=0.4142, simple_loss=0.4349, pruned_loss=0.1967, over 1614472.34 frames. ], batch size: 34, lr: 3.76e-02, grad_scale: 8.0 +2023-02-05 19:21:14,313 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.413e+02 4.263e+02 5.448e+02 6.514e+02 1.215e+03, threshold=1.090e+03, percent-clipped=2.0 +2023-02-05 19:21:42,633 INFO [train.py:901] (1/4) Epoch 1, batch 7350, loss[loss=0.3931, simple_loss=0.4164, pruned_loss=0.1849, over 7800.00 frames. ], tot_loss[loss=0.4137, simple_loss=0.4347, pruned_loss=0.1963, over 1616232.45 frames. ], batch size: 20, lr: 3.75e-02, grad_scale: 8.0 +2023-02-05 19:21:45,493 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.3229, 1.2681, 5.2837, 2.4260, 4.4792, 4.3454, 4.6880, 4.4539], + device='cuda:1'), covar=tensor([0.0256, 0.3443, 0.0155, 0.1169, 0.0725, 0.0292, 0.0251, 0.0371], + device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0338, 0.0178, 0.0212, 0.0223, 0.0201, 0.0173, 0.0202], + device='cuda:1'), out_proj_covar=tensor([9.4270e-05, 1.8591e-04, 1.0904e-04, 1.3403e-04, 1.2752e-04, 1.2063e-04, + 1.0457e-04, 1.2652e-04], device='cuda:1') +2023-02-05 19:21:45,541 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7355.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:21:50,111 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7362.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:21:56,009 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-05 19:22:08,458 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7386.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:22:09,133 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7387.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:22:18,186 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-05 19:22:18,458 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2618, 1.3444, 2.7000, 1.1214, 2.0295, 2.8823, 2.6331, 2.6317], + device='cuda:1'), covar=tensor([0.1769, 0.1865, 0.0405, 0.2383, 0.0793, 0.0290, 0.0309, 0.0430], + device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0265, 0.0163, 0.0256, 0.0190, 0.0128, 0.0126, 0.0185], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-05 19:22:18,974 INFO [train.py:901] (1/4) Epoch 1, batch 7400, loss[loss=0.3528, simple_loss=0.3879, pruned_loss=0.1588, over 7811.00 frames. ], tot_loss[loss=0.4134, simple_loss=0.435, pruned_loss=0.1958, over 1618225.57 frames. ], batch size: 20, lr: 3.74e-02, grad_scale: 8.0 +2023-02-05 19:22:24,409 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.824e+02 4.270e+02 5.603e+02 6.704e+02 2.452e+03, threshold=1.121e+03, percent-clipped=4.0 +2023-02-05 19:22:25,137 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7410.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:22:35,056 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7425.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:22:52,095 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8494, 1.9639, 1.6585, 2.4729, 1.4641, 1.3087, 1.9812, 1.9084], + device='cuda:1'), covar=tensor([0.1053, 0.1340, 0.1434, 0.0417, 0.1692, 0.1937, 0.1468, 0.1159], + device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0327, 0.0311, 0.0200, 0.0343, 0.0347, 0.0389, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0002, 0.0004, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 19:22:52,554 INFO [train.py:901] (1/4) Epoch 1, batch 7450, loss[loss=0.461, simple_loss=0.4653, pruned_loss=0.2284, over 8608.00 frames. ], tot_loss[loss=0.4141, simple_loss=0.4354, pruned_loss=0.1964, over 1613696.01 frames. ], batch size: 39, lr: 3.73e-02, grad_scale: 8.0 +2023-02-05 19:22:56,045 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-05 19:23:00,898 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0647, 3.1525, 2.7466, 1.2231, 2.6538, 2.5949, 2.8211, 2.5175], + device='cuda:1'), covar=tensor([0.1152, 0.0757, 0.1120, 0.3956, 0.0672, 0.0974, 0.1447, 0.0623], + device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0183, 0.0223, 0.0293, 0.0174, 0.0137, 0.0206, 0.0133], + device='cuda:1'), out_proj_covar=tensor([1.9800e-04, 1.2876e-04, 1.4666e-04, 1.8845e-04, 1.1263e-04, 9.9221e-05, + 1.4048e-04, 9.3942e-05], device='cuda:1') +2023-02-05 19:23:02,948 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5026, 4.6919, 4.0454, 1.6778, 3.8114, 3.7858, 4.2559, 3.4266], + device='cuda:1'), covar=tensor([0.0613, 0.0299, 0.0605, 0.3276, 0.0390, 0.0622, 0.0789, 0.0455], + device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0182, 0.0223, 0.0293, 0.0173, 0.0137, 0.0205, 0.0133], + device='cuda:1'), out_proj_covar=tensor([1.9719e-04, 1.2831e-04, 1.4626e-04, 1.8808e-04, 1.1233e-04, 9.9143e-05, + 1.3986e-04, 9.3494e-05], device='cuda:1') +2023-02-05 19:23:05,306 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-05 19:23:10,752 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-05 19:23:27,518 INFO [train.py:901] (1/4) Epoch 1, batch 7500, loss[loss=0.3594, simple_loss=0.3868, pruned_loss=0.166, over 7414.00 frames. ], tot_loss[loss=0.4121, simple_loss=0.4335, pruned_loss=0.1954, over 1610188.30 frames. ], batch size: 17, lr: 3.72e-02, grad_scale: 8.0 +2023-02-05 19:23:34,187 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 4.060e+02 5.044e+02 6.934e+02 1.457e+03, threshold=1.009e+03, percent-clipped=3.0 +2023-02-05 19:23:45,044 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7525.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:23:45,155 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7525.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:24:02,231 INFO [train.py:901] (1/4) Epoch 1, batch 7550, loss[loss=0.4384, simple_loss=0.4721, pruned_loss=0.2023, over 8475.00 frames. ], tot_loss[loss=0.4119, simple_loss=0.4333, pruned_loss=0.1952, over 1606285.46 frames. ], batch size: 29, lr: 3.72e-02, grad_scale: 8.0 +2023-02-05 19:24:36,292 INFO [train.py:901] (1/4) Epoch 1, batch 7600, loss[loss=0.448, simple_loss=0.4561, pruned_loss=0.2199, over 8512.00 frames. ], tot_loss[loss=0.4119, simple_loss=0.4329, pruned_loss=0.1955, over 1606465.61 frames. ], batch size: 28, lr: 3.71e-02, grad_scale: 8.0 +2023-02-05 19:24:41,734 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.765e+02 4.361e+02 5.460e+02 6.853e+02 1.164e+03, threshold=1.092e+03, percent-clipped=2.0 +2023-02-05 19:24:43,938 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7611.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:25:03,348 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7636.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:25:03,854 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7637.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:25:05,945 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7640.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:25:07,275 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7642.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:25:07,354 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7642.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:25:13,185 INFO [train.py:901] (1/4) Epoch 1, batch 7650, loss[loss=0.375, simple_loss=0.3877, pruned_loss=0.1811, over 7725.00 frames. ], tot_loss[loss=0.411, simple_loss=0.4329, pruned_loss=0.1946, over 1610011.06 frames. ], batch size: 18, lr: 3.70e-02, grad_scale: 8.0 +2023-02-05 19:25:23,888 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7667.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:25:46,144 INFO [train.py:901] (1/4) Epoch 1, batch 7700, loss[loss=0.385, simple_loss=0.4094, pruned_loss=0.1803, over 7919.00 frames. ], tot_loss[loss=0.4108, simple_loss=0.4329, pruned_loss=0.1943, over 1609702.38 frames. ], batch size: 20, lr: 3.69e-02, grad_scale: 8.0 +2023-02-05 19:25:51,306 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.517e+02 4.083e+02 4.742e+02 6.161e+02 2.101e+03, threshold=9.483e+02, percent-clipped=6.0 +2023-02-05 19:26:07,597 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-05 19:26:21,658 INFO [train.py:901] (1/4) Epoch 1, batch 7750, loss[loss=0.5122, simple_loss=0.4974, pruned_loss=0.2635, over 7119.00 frames. ], tot_loss[loss=0.4083, simple_loss=0.4303, pruned_loss=0.1931, over 1599073.36 frames. ], batch size: 72, lr: 3.68e-02, grad_scale: 8.0 +2023-02-05 19:26:23,157 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7752.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:26:29,117 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7761.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:26:34,417 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7769.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:26:42,684 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7781.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:26:56,367 INFO [train.py:901] (1/4) Epoch 1, batch 7800, loss[loss=0.4081, simple_loss=0.4242, pruned_loss=0.196, over 8094.00 frames. ], tot_loss[loss=0.4078, simple_loss=0.4299, pruned_loss=0.1928, over 1599798.06 frames. ], batch size: 21, lr: 3.67e-02, grad_scale: 8.0 +2023-02-05 19:26:59,833 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7806.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:27:01,642 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.208e+02 3.720e+02 4.585e+02 5.523e+02 1.290e+03, threshold=9.170e+02, percent-clipped=3.0 +2023-02-05 19:27:09,240 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7820.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:27:29,712 INFO [train.py:901] (1/4) Epoch 1, batch 7850, loss[loss=0.3439, simple_loss=0.376, pruned_loss=0.1558, over 7697.00 frames. ], tot_loss[loss=0.4065, simple_loss=0.4291, pruned_loss=0.192, over 1605266.31 frames. ], batch size: 18, lr: 3.66e-02, grad_scale: 8.0 +2023-02-05 19:27:37,629 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-02-05 19:27:52,021 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7884.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:27:59,883 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7896.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:28:03,027 INFO [train.py:901] (1/4) Epoch 1, batch 7900, loss[loss=0.4361, simple_loss=0.4462, pruned_loss=0.213, over 8471.00 frames. ], tot_loss[loss=0.4067, simple_loss=0.4298, pruned_loss=0.1918, over 1609370.71 frames. ], batch size: 29, lr: 3.66e-02, grad_scale: 8.0 +2023-02-05 19:28:03,947 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0572, 2.1249, 2.9519, 3.2619, 2.7622, 1.8750, 1.9132, 2.2097], + device='cuda:1'), covar=tensor([0.1203, 0.0801, 0.0224, 0.0204, 0.0387, 0.0499, 0.0473, 0.0703], + device='cuda:1'), in_proj_covar=tensor([0.0371, 0.0268, 0.0169, 0.0199, 0.0263, 0.0259, 0.0270, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 19:28:08,438 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.445e+02 3.732e+02 4.923e+02 6.190e+02 1.863e+03, threshold=9.845e+02, percent-clipped=5.0 +2023-02-05 19:28:16,370 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7921.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:28:35,814 INFO [train.py:901] (1/4) Epoch 1, batch 7950, loss[loss=0.395, simple_loss=0.4176, pruned_loss=0.1862, over 8373.00 frames. ], tot_loss[loss=0.4074, simple_loss=0.4301, pruned_loss=0.1924, over 1609618.89 frames. ], batch size: 48, lr: 3.65e-02, grad_scale: 8.0 +2023-02-05 19:28:42,787 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-02-05 19:28:59,118 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7986.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:29:10,050 INFO [train.py:901] (1/4) Epoch 1, batch 8000, loss[loss=0.3893, simple_loss=0.4078, pruned_loss=0.1854, over 7712.00 frames. ], tot_loss[loss=0.4062, simple_loss=0.4289, pruned_loss=0.1917, over 1605448.90 frames. ], batch size: 18, lr: 3.64e-02, grad_scale: 8.0 +2023-02-05 19:29:15,095 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8008.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:29:15,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.650e+02 3.959e+02 4.934e+02 6.403e+02 1.426e+03, threshold=9.868e+02, percent-clipped=4.0 +2023-02-05 19:29:31,421 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8033.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:29:43,004 INFO [train.py:901] (1/4) Epoch 1, batch 8050, loss[loss=0.3727, simple_loss=0.3859, pruned_loss=0.1797, over 7201.00 frames. ], tot_loss[loss=0.4074, simple_loss=0.4288, pruned_loss=0.193, over 1592540.58 frames. ], batch size: 16, lr: 3.63e-02, grad_scale: 16.0 +2023-02-05 19:29:51,668 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0391, 1.3611, 1.5075, 0.3323, 1.3017, 1.1543, 0.1520, 1.5003], + device='cuda:1'), covar=tensor([0.0352, 0.0180, 0.0319, 0.0631, 0.0282, 0.0641, 0.0702, 0.0166], + device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0119, 0.0102, 0.0169, 0.0118, 0.0187, 0.0170, 0.0132], + device='cuda:1'), out_proj_covar=tensor([1.1704e-04, 8.5295e-05, 8.1200e-05, 1.2578e-04, 9.4830e-05, 1.4771e-04, + 1.2942e-04, 9.7166e-05], device='cuda:1') +2023-02-05 19:30:17,053 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-05 19:30:20,856 INFO [train.py:901] (1/4) Epoch 2, batch 0, loss[loss=0.4707, simple_loss=0.464, pruned_loss=0.2387, over 8135.00 frames. ], tot_loss[loss=0.4707, simple_loss=0.464, pruned_loss=0.2387, over 8135.00 frames. ], batch size: 22, lr: 3.56e-02, grad_scale: 8.0 +2023-02-05 19:30:20,856 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 19:30:32,396 INFO [train.py:935] (1/4) Epoch 2, validation: loss=0.3107, simple_loss=0.3861, pruned_loss=0.1176, over 944034.00 frames. +2023-02-05 19:30:32,397 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6470MB +2023-02-05 19:30:36,091 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3882, 1.8338, 3.3712, 2.9442, 2.3144, 1.8939, 1.8951, 2.0937], + device='cuda:1'), covar=tensor([0.0806, 0.0772, 0.0110, 0.0208, 0.0392, 0.0425, 0.0501, 0.0609], + device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0275, 0.0175, 0.0200, 0.0269, 0.0262, 0.0282, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 19:30:44,057 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8101.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:30:46,380 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-05 19:30:46,617 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-05 19:30:46,677 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8105.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:30:49,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.305e+02 3.846e+02 4.676e+02 6.027e+02 1.450e+03, threshold=9.352e+02, percent-clipped=5.0 +2023-02-05 19:31:06,745 INFO [train.py:901] (1/4) Epoch 2, batch 50, loss[loss=0.3337, simple_loss=0.38, pruned_loss=0.1437, over 7791.00 frames. ], tot_loss[loss=0.4019, simple_loss=0.4268, pruned_loss=0.1885, over 359871.00 frames. ], batch size: 19, lr: 3.55e-02, grad_scale: 8.0 +2023-02-05 19:31:11,120 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8140.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:31:20,804 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-05 19:31:28,339 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8164.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:31:29,157 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8165.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:31:41,593 INFO [train.py:901] (1/4) Epoch 2, batch 100, loss[loss=0.4516, simple_loss=0.4631, pruned_loss=0.22, over 7661.00 frames. ], tot_loss[loss=0.4127, simple_loss=0.4354, pruned_loss=0.195, over 640604.10 frames. ], batch size: 19, lr: 3.54e-02, grad_scale: 8.0 +2023-02-05 19:31:44,274 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-05 19:31:59,424 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.305e+02 4.246e+02 4.943e+02 6.491e+02 9.375e+02, threshold=9.885e+02, percent-clipped=1.0 +2023-02-05 19:32:06,340 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:32:07,760 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0998, 2.4838, 4.6106, 1.1780, 3.0390, 2.3056, 1.7895, 2.5047], + device='cuda:1'), covar=tensor([0.0963, 0.1161, 0.0281, 0.1640, 0.0988, 0.1419, 0.0902, 0.1387], + device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0316, 0.0312, 0.0356, 0.0403, 0.0372, 0.0324, 0.0388], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 19:32:15,455 INFO [train.py:901] (1/4) Epoch 2, batch 150, loss[loss=0.3442, simple_loss=0.384, pruned_loss=0.1523, over 8031.00 frames. ], tot_loss[loss=0.4053, simple_loss=0.4303, pruned_loss=0.1901, over 858424.97 frames. ], batch size: 22, lr: 3.53e-02, grad_scale: 8.0 +2023-02-05 19:32:47,791 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8279.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:32:50,388 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8283.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:32:50,923 INFO [train.py:901] (1/4) Epoch 2, batch 200, loss[loss=0.4083, simple_loss=0.4282, pruned_loss=0.1942, over 7976.00 frames. ], tot_loss[loss=0.405, simple_loss=0.4302, pruned_loss=0.1899, over 1028462.56 frames. ], batch size: 21, lr: 3.52e-02, grad_scale: 8.0 +2023-02-05 19:32:53,835 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-02-05 19:33:08,598 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.581e+02 3.727e+02 4.975e+02 6.903e+02 1.681e+03, threshold=9.950e+02, percent-clipped=7.0 +2023-02-05 19:33:24,848 INFO [train.py:901] (1/4) Epoch 2, batch 250, loss[loss=0.4791, simple_loss=0.4891, pruned_loss=0.2346, over 8450.00 frames. ], tot_loss[loss=0.405, simple_loss=0.43, pruned_loss=0.19, over 1157959.05 frames. ], batch size: 27, lr: 3.52e-02, grad_scale: 8.0 +2023-02-05 19:33:27,774 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9244, 1.5722, 1.3958, 1.4134, 1.7887, 1.5096, 1.4375, 1.7570], + device='cuda:1'), covar=tensor([0.1255, 0.1939, 0.2697, 0.2145, 0.1106, 0.1904, 0.1485, 0.1188], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0281, 0.0299, 0.0272, 0.0260, 0.0252, 0.0257, 0.0246], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-02-05 19:33:36,310 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-05 19:33:40,681 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8357.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:33:46,006 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-05 19:33:58,447 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8382.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:33:59,607 INFO [train.py:901] (1/4) Epoch 2, batch 300, loss[loss=0.4504, simple_loss=0.4706, pruned_loss=0.2151, over 8462.00 frames. ], tot_loss[loss=0.4029, simple_loss=0.4282, pruned_loss=0.1888, over 1258159.82 frames. ], batch size: 27, lr: 3.51e-02, grad_scale: 8.0 +2023-02-05 19:34:18,663 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 4.043e+02 4.737e+02 5.583e+02 9.957e+02, threshold=9.474e+02, percent-clipped=1.0 +2023-02-05 19:34:29,133 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 +2023-02-05 19:34:35,499 INFO [train.py:901] (1/4) Epoch 2, batch 350, loss[loss=0.3569, simple_loss=0.3946, pruned_loss=0.1596, over 8099.00 frames. ], tot_loss[loss=0.403, simple_loss=0.4282, pruned_loss=0.1889, over 1338979.22 frames. ], batch size: 23, lr: 3.50e-02, grad_scale: 8.0 +2023-02-05 19:35:00,611 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-02-05 19:35:03,558 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8476.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:35:07,681 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.8351, 1.1315, 0.9131, 0.7851, 0.7160, 1.0305, 0.0144, 0.6527], + device='cuda:1'), covar=tensor([0.0588, 0.0454, 0.0366, 0.0496, 0.0562, 0.0342, 0.1518, 0.0819], + device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0118, 0.0101, 0.0148, 0.0119, 0.0091, 0.0167, 0.0143], + device='cuda:1'), out_proj_covar=tensor([1.1985e-04, 1.1085e-04, 9.0684e-05, 1.2712e-04, 1.1273e-04, 8.3598e-05, + 1.4563e-04, 1.2831e-04], device='cuda:1') +2023-02-05 19:35:09,448 INFO [train.py:901] (1/4) Epoch 2, batch 400, loss[loss=0.3996, simple_loss=0.4211, pruned_loss=0.189, over 8125.00 frames. ], tot_loss[loss=0.4021, simple_loss=0.4275, pruned_loss=0.1883, over 1401681.04 frames. ], batch size: 22, lr: 3.49e-02, grad_scale: 8.0 +2023-02-05 19:35:20,907 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8501.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:35:27,437 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.659e+02 4.339e+02 4.887e+02 6.099e+02 1.134e+03, threshold=9.773e+02, percent-clipped=6.0 +2023-02-05 19:35:43,490 INFO [train.py:901] (1/4) Epoch 2, batch 450, loss[loss=0.4188, simple_loss=0.4422, pruned_loss=0.1977, over 8334.00 frames. ], tot_loss[loss=0.4043, simple_loss=0.4292, pruned_loss=0.1897, over 1447416.92 frames. ], batch size: 26, lr: 3.49e-02, grad_scale: 8.0 +2023-02-05 19:35:44,346 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8535.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:36:01,885 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8560.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:36:18,004 INFO [train.py:901] (1/4) Epoch 2, batch 500, loss[loss=0.375, simple_loss=0.4124, pruned_loss=0.1688, over 7972.00 frames. ], tot_loss[loss=0.4001, simple_loss=0.4263, pruned_loss=0.187, over 1485513.68 frames. ], batch size: 21, lr: 3.48e-02, grad_scale: 8.0 +2023-02-05 19:36:36,156 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.910e+02 4.803e+02 5.619e+02 9.699e+02, threshold=9.605e+02, percent-clipped=0.0 +2023-02-05 19:36:47,558 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8627.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:36:52,677 INFO [train.py:901] (1/4) Epoch 2, batch 550, loss[loss=0.4441, simple_loss=0.4553, pruned_loss=0.2165, over 8287.00 frames. ], tot_loss[loss=0.3993, simple_loss=0.4252, pruned_loss=0.1867, over 1510781.78 frames. ], batch size: 23, lr: 3.47e-02, grad_scale: 8.0 +2023-02-05 19:37:26,531 INFO [train.py:901] (1/4) Epoch 2, batch 600, loss[loss=0.4042, simple_loss=0.4015, pruned_loss=0.2034, over 7439.00 frames. ], tot_loss[loss=0.3988, simple_loss=0.4249, pruned_loss=0.1864, over 1535830.68 frames. ], batch size: 17, lr: 3.46e-02, grad_scale: 8.0 +2023-02-05 19:37:29,013 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-05 19:37:32,131 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5103, 1.6727, 1.6688, 2.3604, 0.9514, 1.1686, 1.5089, 1.7280], + device='cuda:1'), covar=tensor([0.1148, 0.1385, 0.1158, 0.0374, 0.1972, 0.1995, 0.1733, 0.1173], + device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0326, 0.0315, 0.0208, 0.0341, 0.0351, 0.0385, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-02-05 19:37:43,318 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.752e+02 3.934e+02 5.073e+02 6.758e+02 1.500e+03, threshold=1.015e+03, percent-clipped=5.0 +2023-02-05 19:37:44,753 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-05 19:37:59,728 INFO [train.py:901] (1/4) Epoch 2, batch 650, loss[loss=0.4536, simple_loss=0.4735, pruned_loss=0.2168, over 8203.00 frames. ], tot_loss[loss=0.3998, simple_loss=0.4257, pruned_loss=0.187, over 1551972.71 frames. ], batch size: 23, lr: 3.46e-02, grad_scale: 8.0 +2023-02-05 19:38:05,386 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8742.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:38:31,187 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8778.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:38:31,223 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3800, 1.3703, 1.3335, 1.1313, 1.5391, 1.2475, 1.0795, 1.4982], + device='cuda:1'), covar=tensor([0.1289, 0.1994, 0.2307, 0.2283, 0.0932, 0.1962, 0.1413, 0.1009], + device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0278, 0.0292, 0.0278, 0.0258, 0.0255, 0.0258, 0.0245], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-02-05 19:38:35,550 INFO [train.py:901] (1/4) Epoch 2, batch 700, loss[loss=0.3339, simple_loss=0.3738, pruned_loss=0.147, over 8054.00 frames. ], tot_loss[loss=0.3991, simple_loss=0.4251, pruned_loss=0.1866, over 1569840.40 frames. ], batch size: 20, lr: 3.45e-02, grad_scale: 8.0 +2023-02-05 19:38:53,112 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.421e+02 3.759e+02 4.676e+02 6.060e+02 1.461e+03, threshold=9.352e+02, percent-clipped=1.0 +2023-02-05 19:39:09,181 INFO [train.py:901] (1/4) Epoch 2, batch 750, loss[loss=0.4373, simple_loss=0.4575, pruned_loss=0.2086, over 8465.00 frames. ], tot_loss[loss=0.3976, simple_loss=0.4242, pruned_loss=0.1855, over 1579481.82 frames. ], batch size: 27, lr: 3.44e-02, grad_scale: 8.0 +2023-02-05 19:39:23,216 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4328, 2.1749, 2.3622, 0.6779, 2.4197, 1.6001, 1.0912, 1.5649], + device='cuda:1'), covar=tensor([0.0419, 0.0162, 0.0232, 0.0699, 0.0256, 0.0528, 0.0717, 0.0345], + device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0126, 0.0107, 0.0174, 0.0119, 0.0200, 0.0180, 0.0145], + device='cuda:1'), out_proj_covar=tensor([1.1862e-04, 8.9553e-05, 8.2196e-05, 1.2736e-04, 9.1641e-05, 1.5391e-04, + 1.3355e-04, 1.0755e-04], device='cuda:1') +2023-02-05 19:39:26,415 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-05 19:39:35,548 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-05 19:39:44,342 INFO [train.py:901] (1/4) Epoch 2, batch 800, loss[loss=0.4073, simple_loss=0.4445, pruned_loss=0.1851, over 8334.00 frames. ], tot_loss[loss=0.4016, simple_loss=0.4268, pruned_loss=0.1882, over 1589496.35 frames. ], batch size: 26, lr: 3.43e-02, grad_scale: 8.0 +2023-02-05 19:40:02,285 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.406e+02 4.043e+02 5.225e+02 6.708e+02 1.302e+03, threshold=1.045e+03, percent-clipped=9.0 +2023-02-05 19:40:18,497 INFO [train.py:901] (1/4) Epoch 2, batch 850, loss[loss=0.4011, simple_loss=0.4144, pruned_loss=0.1939, over 7710.00 frames. ], tot_loss[loss=0.3975, simple_loss=0.4241, pruned_loss=0.1854, over 1596321.47 frames. ], batch size: 18, lr: 3.43e-02, grad_scale: 8.0 +2023-02-05 19:40:26,055 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8945.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:40:52,653 INFO [train.py:901] (1/4) Epoch 2, batch 900, loss[loss=0.4227, simple_loss=0.4601, pruned_loss=0.1927, over 8723.00 frames. ], tot_loss[loss=0.3956, simple_loss=0.4233, pruned_loss=0.1839, over 1604426.31 frames. ], batch size: 30, lr: 3.42e-02, grad_scale: 8.0 +2023-02-05 19:41:03,946 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8998.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:41:08,138 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9004.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:41:12,012 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.317e+02 3.660e+02 4.402e+02 6.333e+02 1.420e+03, threshold=8.805e+02, percent-clipped=4.0 +2023-02-05 19:41:21,856 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9023.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:41:27,232 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9031.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:41:29,111 INFO [train.py:901] (1/4) Epoch 2, batch 950, loss[loss=0.4026, simple_loss=0.4033, pruned_loss=0.201, over 7927.00 frames. ], tot_loss[loss=0.3954, simple_loss=0.4231, pruned_loss=0.1838, over 1605895.35 frames. ], batch size: 20, lr: 3.41e-02, grad_scale: 8.0 +2023-02-05 19:41:57,087 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-05 19:42:04,019 INFO [train.py:901] (1/4) Epoch 2, batch 1000, loss[loss=0.3815, simple_loss=0.4066, pruned_loss=0.1782, over 8547.00 frames. ], tot_loss[loss=0.3936, simple_loss=0.4219, pruned_loss=0.1826, over 1609324.99 frames. ], batch size: 31, lr: 3.40e-02, grad_scale: 8.0 +2023-02-05 19:42:21,659 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-05 19:42:22,620 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.505e+02 3.676e+02 4.681e+02 5.718e+02 9.745e+02, threshold=9.362e+02, percent-clipped=2.0 +2023-02-05 19:42:30,650 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9122.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:42:31,271 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-05 19:42:39,155 INFO [train.py:901] (1/4) Epoch 2, batch 1050, loss[loss=0.4249, simple_loss=0.4431, pruned_loss=0.2033, over 8465.00 frames. ], tot_loss[loss=0.3929, simple_loss=0.4213, pruned_loss=0.1822, over 1610697.99 frames. ], batch size: 25, lr: 3.40e-02, grad_scale: 8.0 +2023-02-05 19:42:39,685 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-05 19:42:43,224 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-05 19:43:12,162 INFO [train.py:901] (1/4) Epoch 2, batch 1100, loss[loss=0.368, simple_loss=0.3942, pruned_loss=0.1709, over 7276.00 frames. ], tot_loss[loss=0.3925, simple_loss=0.4207, pruned_loss=0.1822, over 1607257.45 frames. ], batch size: 16, lr: 3.39e-02, grad_scale: 8.0 +2023-02-05 19:43:30,057 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.440e+02 3.895e+02 4.986e+02 6.293e+02 1.172e+03, threshold=9.973e+02, percent-clipped=2.0 +2023-02-05 19:43:47,492 INFO [train.py:901] (1/4) Epoch 2, batch 1150, loss[loss=0.4165, simple_loss=0.4539, pruned_loss=0.1895, over 8189.00 frames. ], tot_loss[loss=0.3935, simple_loss=0.4211, pruned_loss=0.1829, over 1610358.05 frames. ], batch size: 23, lr: 3.38e-02, grad_scale: 8.0 +2023-02-05 19:43:49,760 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9237.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:43:50,987 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-05 19:44:08,368 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-02-05 19:44:22,137 INFO [train.py:901] (1/4) Epoch 2, batch 1200, loss[loss=0.4109, simple_loss=0.4377, pruned_loss=0.192, over 8591.00 frames. ], tot_loss[loss=0.3924, simple_loss=0.421, pruned_loss=0.1819, over 1614361.03 frames. ], batch size: 39, lr: 3.38e-02, grad_scale: 8.0 +2023-02-05 19:44:25,541 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9289.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:44:41,016 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.810e+02 4.160e+02 4.885e+02 6.720e+02 4.965e+03, threshold=9.769e+02, percent-clipped=5.0 +2023-02-05 19:44:41,470 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-05 19:44:56,724 INFO [train.py:901] (1/4) Epoch 2, batch 1250, loss[loss=0.4003, simple_loss=0.4367, pruned_loss=0.182, over 8242.00 frames. ], tot_loss[loss=0.3933, simple_loss=0.4221, pruned_loss=0.1823, over 1618951.73 frames. ], batch size: 24, lr: 3.37e-02, grad_scale: 4.0 +2023-02-05 19:45:02,246 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2887, 1.5177, 2.4095, 1.0117, 1.7094, 1.4698, 1.3779, 1.5512], + device='cuda:1'), covar=tensor([0.1075, 0.1057, 0.0351, 0.1568, 0.0919, 0.1495, 0.0970, 0.0986], + device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0329, 0.0349, 0.0373, 0.0426, 0.0395, 0.0339, 0.0413], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 19:45:07,485 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9348.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:45:17,413 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3418, 1.0956, 2.9560, 1.1115, 1.8708, 3.3369, 3.1569, 2.8928], + device='cuda:1'), covar=tensor([0.1795, 0.2184, 0.0545, 0.2728, 0.1069, 0.0380, 0.0407, 0.0782], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0255, 0.0177, 0.0251, 0.0185, 0.0138, 0.0137, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-05 19:45:25,821 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9375.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:45:27,212 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0701, 3.1914, 2.7951, 1.4825, 2.5670, 2.6918, 2.8057, 2.4230], + device='cuda:1'), covar=tensor([0.1154, 0.0811, 0.1059, 0.3605, 0.0758, 0.0660, 0.1432, 0.0660], + device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0198, 0.0239, 0.0309, 0.0201, 0.0151, 0.0218, 0.0147], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 19:45:31,819 INFO [train.py:901] (1/4) Epoch 2, batch 1300, loss[loss=0.3774, simple_loss=0.414, pruned_loss=0.1704, over 8347.00 frames. ], tot_loss[loss=0.3926, simple_loss=0.4216, pruned_loss=0.1818, over 1618919.74 frames. ], batch size: 24, lr: 3.36e-02, grad_scale: 4.0 +2023-02-05 19:45:45,162 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0014, 1.8622, 3.1822, 2.8239, 2.5282, 1.8380, 1.5262, 1.9701], + device='cuda:1'), covar=tensor([0.1108, 0.0991, 0.0180, 0.0276, 0.0407, 0.0467, 0.0607, 0.0761], + device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0337, 0.0230, 0.0268, 0.0355, 0.0309, 0.0335, 0.0383], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 19:45:45,735 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9404.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:45:50,305 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.161e+02 4.162e+02 5.656e+02 7.688e+02 2.529e+03, threshold=1.131e+03, percent-clipped=11.0 +2023-02-05 19:45:57,157 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7858, 6.1501, 5.2296, 1.7613, 4.6430, 5.2377, 5.5275, 4.6403], + device='cuda:1'), covar=tensor([0.0842, 0.0295, 0.0589, 0.4589, 0.0423, 0.0490, 0.0953, 0.0509], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0201, 0.0245, 0.0316, 0.0201, 0.0156, 0.0218, 0.0152], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 19:46:05,038 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9432.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:46:06,286 INFO [train.py:901] (1/4) Epoch 2, batch 1350, loss[loss=0.3793, simple_loss=0.417, pruned_loss=0.1708, over 8511.00 frames. ], tot_loss[loss=0.3918, simple_loss=0.4207, pruned_loss=0.1815, over 1616361.25 frames. ], batch size: 26, lr: 3.36e-02, grad_scale: 4.0 +2023-02-05 19:46:27,274 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9463.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:46:27,429 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.89 vs. limit=5.0 +2023-02-05 19:46:41,365 INFO [train.py:901] (1/4) Epoch 2, batch 1400, loss[loss=0.4181, simple_loss=0.4548, pruned_loss=0.1907, over 8390.00 frames. ], tot_loss[loss=0.3922, simple_loss=0.4207, pruned_loss=0.1818, over 1616591.56 frames. ], batch size: 49, lr: 3.35e-02, grad_scale: 4.0 +2023-02-05 19:46:45,509 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9490.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:46:47,583 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9493.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:46:56,270 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8411, 0.9313, 4.1050, 1.7771, 2.0587, 4.8963, 4.2167, 4.4501], + device='cuda:1'), covar=tensor([0.1615, 0.2334, 0.0277, 0.2044, 0.1039, 0.0183, 0.0237, 0.0346], + device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0266, 0.0180, 0.0254, 0.0193, 0.0144, 0.0142, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-05 19:46:59,487 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.192e+02 3.889e+02 4.981e+02 6.326e+02 1.555e+03, threshold=9.962e+02, percent-clipped=1.0 +2023-02-05 19:47:04,250 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9518.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:47:15,072 INFO [train.py:901] (1/4) Epoch 2, batch 1450, loss[loss=0.4216, simple_loss=0.4394, pruned_loss=0.2019, over 8337.00 frames. ], tot_loss[loss=0.3929, simple_loss=0.421, pruned_loss=0.1824, over 1617799.11 frames. ], batch size: 26, lr: 3.34e-02, grad_scale: 4.0 +2023-02-05 19:47:19,043 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-05 19:47:49,259 INFO [train.py:901] (1/4) Epoch 2, batch 1500, loss[loss=0.4998, simple_loss=0.4998, pruned_loss=0.2499, over 8610.00 frames. ], tot_loss[loss=0.3909, simple_loss=0.4196, pruned_loss=0.1811, over 1618494.21 frames. ], batch size: 31, lr: 3.33e-02, grad_scale: 4.0 +2023-02-05 19:48:01,350 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9602.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:48:07,902 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.496e+02 4.006e+02 4.905e+02 6.157e+02 1.300e+03, threshold=9.811e+02, percent-clipped=3.0 +2023-02-05 19:48:23,391 INFO [train.py:901] (1/4) Epoch 2, batch 1550, loss[loss=0.372, simple_loss=0.4131, pruned_loss=0.1654, over 8447.00 frames. ], tot_loss[loss=0.3903, simple_loss=0.4193, pruned_loss=0.1806, over 1617952.08 frames. ], batch size: 27, lr: 3.33e-02, grad_scale: 4.0 +2023-02-05 19:48:41,695 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9660.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:48:52,347 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9676.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:48:57,573 INFO [train.py:901] (1/4) Epoch 2, batch 1600, loss[loss=0.3975, simple_loss=0.4261, pruned_loss=0.1845, over 8664.00 frames. ], tot_loss[loss=0.39, simple_loss=0.4189, pruned_loss=0.1806, over 1617414.17 frames. ], batch size: 39, lr: 3.32e-02, grad_scale: 8.0 +2023-02-05 19:48:58,394 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9685.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:17,081 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.659e+02 4.192e+02 5.177e+02 6.492e+02 1.266e+03, threshold=1.035e+03, percent-clipped=2.0 +2023-02-05 19:49:22,896 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:33,628 INFO [train.py:901] (1/4) Epoch 2, batch 1650, loss[loss=0.4183, simple_loss=0.4457, pruned_loss=0.1955, over 8470.00 frames. ], tot_loss[loss=0.391, simple_loss=0.4193, pruned_loss=0.1813, over 1619802.02 frames. ], batch size: 29, lr: 3.31e-02, grad_scale: 8.0 +2023-02-05 19:49:40,258 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9744.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:41,550 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9746.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:51,527 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9761.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:58,954 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9771.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:50:02,238 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9776.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:50:07,341 INFO [train.py:901] (1/4) Epoch 2, batch 1700, loss[loss=0.3455, simple_loss=0.3913, pruned_loss=0.1498, over 7978.00 frames. ], tot_loss[loss=0.3912, simple_loss=0.4194, pruned_loss=0.1815, over 1621133.19 frames. ], batch size: 21, lr: 3.31e-02, grad_scale: 8.0 +2023-02-05 19:50:26,240 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.640e+02 4.068e+02 5.098e+02 6.535e+02 1.207e+03, threshold=1.020e+03, percent-clipped=5.0 +2023-02-05 19:50:42,247 INFO [train.py:901] (1/4) Epoch 2, batch 1750, loss[loss=0.3973, simple_loss=0.4266, pruned_loss=0.184, over 8245.00 frames. ], tot_loss[loss=0.3894, simple_loss=0.4183, pruned_loss=0.1802, over 1619032.98 frames. ], batch size: 22, lr: 3.30e-02, grad_scale: 8.0 +2023-02-05 19:51:16,321 INFO [train.py:901] (1/4) Epoch 2, batch 1800, loss[loss=0.3697, simple_loss=0.402, pruned_loss=0.1686, over 8025.00 frames. ], tot_loss[loss=0.3905, simple_loss=0.4191, pruned_loss=0.1809, over 1618907.09 frames. ], batch size: 22, lr: 3.29e-02, grad_scale: 8.0 +2023-02-05 19:51:21,271 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9891.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:51:34,076 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.365e+02 4.111e+02 5.198e+02 6.626e+02 1.120e+03, threshold=1.040e+03, percent-clipped=3.0 +2023-02-05 19:51:49,949 INFO [train.py:901] (1/4) Epoch 2, batch 1850, loss[loss=0.3244, simple_loss=0.373, pruned_loss=0.1379, over 7419.00 frames. ], tot_loss[loss=0.3908, simple_loss=0.4195, pruned_loss=0.1811, over 1618957.19 frames. ], batch size: 17, lr: 3.29e-02, grad_scale: 8.0 +2023-02-05 19:51:51,600 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-02-05 19:51:58,651 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9946.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:52:11,388 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3374, 1.7804, 1.4396, 1.1696, 2.1074, 1.6052, 1.8790, 2.2281], + device='cuda:1'), covar=tensor([0.1141, 0.1926, 0.2529, 0.2158, 0.0976, 0.2182, 0.1274, 0.0861], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0269, 0.0290, 0.0260, 0.0240, 0.0252, 0.0235, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-02-05 19:52:24,354 INFO [train.py:901] (1/4) Epoch 2, batch 1900, loss[loss=0.3323, simple_loss=0.3846, pruned_loss=0.14, over 8357.00 frames. ], tot_loss[loss=0.3879, simple_loss=0.4175, pruned_loss=0.1791, over 1615913.55 frames. ], batch size: 24, lr: 3.28e-02, grad_scale: 8.0 +2023-02-05 19:52:43,808 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.445e+02 3.513e+02 4.327e+02 5.785e+02 1.080e+03, threshold=8.653e+02, percent-clipped=1.0 +2023-02-05 19:52:49,914 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10020.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:52:54,613 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-05 19:52:55,762 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.68 vs. limit=5.0 +2023-02-05 19:52:59,200 INFO [train.py:901] (1/4) Epoch 2, batch 1950, loss[loss=0.4162, simple_loss=0.4464, pruned_loss=0.193, over 8517.00 frames. ], tot_loss[loss=0.3865, simple_loss=0.4172, pruned_loss=0.1779, over 1616026.25 frames. ], batch size: 28, lr: 3.27e-02, grad_scale: 8.0 +2023-02-05 19:53:06,846 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-05 19:53:15,909 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10057.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:53:19,402 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10061.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:53:25,604 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-05 19:53:33,050 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10080.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:53:35,465 INFO [train.py:901] (1/4) Epoch 2, batch 2000, loss[loss=0.4084, simple_loss=0.4202, pruned_loss=0.1983, over 7930.00 frames. ], tot_loss[loss=0.3862, simple_loss=0.4169, pruned_loss=0.1778, over 1616185.74 frames. ], batch size: 20, lr: 3.27e-02, grad_scale: 8.0 +2023-02-05 19:53:50,417 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10105.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:53:55,728 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.648e+02 4.167e+02 5.413e+02 6.926e+02 6.671e+03, threshold=1.083e+03, percent-clipped=14.0 +2023-02-05 19:54:09,409 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3152, 1.6495, 1.2195, 1.0694, 2.0348, 1.3153, 1.3997, 2.0868], + device='cuda:1'), covar=tensor([0.1060, 0.1879, 0.2671, 0.2380, 0.1045, 0.2179, 0.1396, 0.0851], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0263, 0.0281, 0.0260, 0.0236, 0.0246, 0.0230, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-02-05 19:54:10,560 INFO [train.py:901] (1/4) Epoch 2, batch 2050, loss[loss=0.3689, simple_loss=0.3997, pruned_loss=0.1691, over 7521.00 frames. ], tot_loss[loss=0.3854, simple_loss=0.4162, pruned_loss=0.1773, over 1617248.95 frames. ], batch size: 18, lr: 3.26e-02, grad_scale: 4.0 +2023-02-05 19:54:11,460 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10135.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:54:19,453 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10147.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:54:36,896 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10172.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:54:45,482 INFO [train.py:901] (1/4) Epoch 2, batch 2100, loss[loss=0.3311, simple_loss=0.3852, pruned_loss=0.1385, over 8343.00 frames. ], tot_loss[loss=0.384, simple_loss=0.4148, pruned_loss=0.1766, over 1613909.58 frames. ], batch size: 25, lr: 3.25e-02, grad_scale: 4.0 +2023-02-05 19:55:06,155 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.637e+02 3.788e+02 4.646e+02 5.840e+02 1.328e+03, threshold=9.292e+02, percent-clipped=3.0 +2023-02-05 19:55:09,447 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-05 19:55:11,269 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:55:20,252 INFO [train.py:901] (1/4) Epoch 2, batch 2150, loss[loss=0.3521, simple_loss=0.3903, pruned_loss=0.157, over 8103.00 frames. ], tot_loss[loss=0.3866, simple_loss=0.4169, pruned_loss=0.1781, over 1615094.52 frames. ], batch size: 23, lr: 3.25e-02, grad_scale: 4.0 +2023-02-05 19:55:53,988 INFO [train.py:901] (1/4) Epoch 2, batch 2200, loss[loss=0.4029, simple_loss=0.4344, pruned_loss=0.1857, over 8255.00 frames. ], tot_loss[loss=0.3864, simple_loss=0.4166, pruned_loss=0.1781, over 1614106.32 frames. ], batch size: 24, lr: 3.24e-02, grad_scale: 4.0 +2023-02-05 19:56:13,538 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1197, 2.1497, 3.8391, 4.0579, 3.1251, 1.6851, 1.1982, 1.9442], + device='cuda:1'), covar=tensor([0.1607, 0.1352, 0.0181, 0.0284, 0.0463, 0.0918, 0.1478, 0.1230], + device='cuda:1'), in_proj_covar=tensor([0.0445, 0.0351, 0.0252, 0.0291, 0.0373, 0.0328, 0.0347, 0.0397], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 19:56:14,586 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 3.803e+02 4.971e+02 6.310e+02 1.458e+03, threshold=9.942e+02, percent-clipped=6.0 +2023-02-05 19:56:18,166 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10317.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:56:29,273 INFO [train.py:901] (1/4) Epoch 2, batch 2250, loss[loss=0.3922, simple_loss=0.4307, pruned_loss=0.1769, over 8515.00 frames. ], tot_loss[loss=0.3868, simple_loss=0.4169, pruned_loss=0.1783, over 1615999.13 frames. ], batch size: 28, lr: 3.24e-02, grad_scale: 4.0 +2023-02-05 19:56:34,611 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10342.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:56:37,306 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6564, 1.0548, 3.1759, 1.1651, 1.9088, 3.8937, 3.4563, 3.3619], + device='cuda:1'), covar=tensor([0.1521, 0.1880, 0.0384, 0.2120, 0.0906, 0.0172, 0.0219, 0.0410], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0261, 0.0175, 0.0248, 0.0191, 0.0149, 0.0136, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-05 19:57:03,222 INFO [train.py:901] (1/4) Epoch 2, batch 2300, loss[loss=0.3645, simple_loss=0.4035, pruned_loss=0.1627, over 8508.00 frames. ], tot_loss[loss=0.3869, simple_loss=0.4174, pruned_loss=0.1782, over 1617265.54 frames. ], batch size: 49, lr: 3.23e-02, grad_scale: 4.0 +2023-02-05 19:57:08,303 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10391.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:57:15,021 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10401.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:57:23,819 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.389e+02 3.989e+02 5.161e+02 7.086e+02 1.471e+03, threshold=1.032e+03, percent-clipped=7.0 +2023-02-05 19:57:25,925 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10416.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:57:31,794 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10424.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:57:33,875 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10427.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:57:39,145 INFO [train.py:901] (1/4) Epoch 2, batch 2350, loss[loss=0.3223, simple_loss=0.3792, pruned_loss=0.1327, over 7787.00 frames. ], tot_loss[loss=0.3859, simple_loss=0.4169, pruned_loss=0.1774, over 1619616.40 frames. ], batch size: 19, lr: 3.22e-02, grad_scale: 4.0 +2023-02-05 19:58:05,152 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10472.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:58:07,803 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10476.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:58:12,898 INFO [train.py:901] (1/4) Epoch 2, batch 2400, loss[loss=0.4567, simple_loss=0.4688, pruned_loss=0.2223, over 8465.00 frames. ], tot_loss[loss=0.384, simple_loss=0.4154, pruned_loss=0.1763, over 1615023.31 frames. ], batch size: 25, lr: 3.22e-02, grad_scale: 8.0 +2023-02-05 19:58:24,664 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10501.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:58:32,501 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.364e+02 3.956e+02 5.047e+02 6.263e+02 1.564e+03, threshold=1.009e+03, percent-clipped=2.0 +2023-02-05 19:58:34,737 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10516.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:58:36,282 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-05 19:58:47,320 INFO [train.py:901] (1/4) Epoch 2, batch 2450, loss[loss=0.3462, simple_loss=0.396, pruned_loss=0.1482, over 8498.00 frames. ], tot_loss[loss=0.385, simple_loss=0.4155, pruned_loss=0.1772, over 1610869.61 frames. ], batch size: 26, lr: 3.21e-02, grad_scale: 8.0 +2023-02-05 19:58:50,861 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10539.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:59:22,252 INFO [train.py:901] (1/4) Epoch 2, batch 2500, loss[loss=0.36, simple_loss=0.4057, pruned_loss=0.1571, over 8468.00 frames. ], tot_loss[loss=0.3834, simple_loss=0.4143, pruned_loss=0.1762, over 1610464.58 frames. ], batch size: 29, lr: 3.20e-02, grad_scale: 8.0 +2023-02-05 19:59:42,162 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.520e+02 3.522e+02 4.438e+02 6.473e+02 1.354e+03, threshold=8.876e+02, percent-clipped=4.0 +2023-02-05 19:59:53,425 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5997, 1.8099, 1.4814, 1.1078, 2.1371, 1.5928, 1.8902, 2.0854], + device='cuda:1'), covar=tensor([0.1051, 0.1740, 0.2333, 0.2241, 0.0880, 0.1871, 0.1229, 0.0897], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0263, 0.0290, 0.0256, 0.0235, 0.0250, 0.0229, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-02-05 19:59:55,950 INFO [train.py:901] (1/4) Epoch 2, batch 2550, loss[loss=0.384, simple_loss=0.4225, pruned_loss=0.1728, over 8561.00 frames. ], tot_loss[loss=0.382, simple_loss=0.4132, pruned_loss=0.1754, over 1613899.48 frames. ], batch size: 31, lr: 3.20e-02, grad_scale: 8.0 +2023-02-05 19:59:58,132 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1268, 0.8819, 3.1557, 0.8946, 2.6406, 2.6291, 2.7390, 2.7955], + device='cuda:1'), covar=tensor([0.0340, 0.2829, 0.0396, 0.1538, 0.1118, 0.0470, 0.0371, 0.0439], + device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0357, 0.0213, 0.0242, 0.0290, 0.0229, 0.0213, 0.0246], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 20:00:19,397 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.3119, 2.5962, 4.0163, 4.2184, 2.8761, 2.5072, 1.9614, 2.2593], + device='cuda:1'), covar=tensor([0.0720, 0.0932, 0.0180, 0.0199, 0.0434, 0.0344, 0.0476, 0.0865], + device='cuda:1'), in_proj_covar=tensor([0.0467, 0.0370, 0.0256, 0.0298, 0.0377, 0.0335, 0.0358, 0.0403], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:00:31,356 INFO [train.py:901] (1/4) Epoch 2, batch 2600, loss[loss=0.3594, simple_loss=0.3992, pruned_loss=0.1598, over 8472.00 frames. ], tot_loss[loss=0.381, simple_loss=0.4127, pruned_loss=0.1746, over 1619873.40 frames. ], batch size: 25, lr: 3.19e-02, grad_scale: 8.0 +2023-02-05 20:00:50,499 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 4.188e+02 4.914e+02 6.333e+02 1.432e+03, threshold=9.828e+02, percent-clipped=6.0 +2023-02-05 20:01:05,101 INFO [train.py:901] (1/4) Epoch 2, batch 2650, loss[loss=0.3412, simple_loss=0.3835, pruned_loss=0.1494, over 8033.00 frames. ], tot_loss[loss=0.3821, simple_loss=0.4139, pruned_loss=0.1751, over 1624235.76 frames. ], batch size: 22, lr: 3.19e-02, grad_scale: 8.0 +2023-02-05 20:01:19,079 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-05 20:01:24,207 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10762.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:01:30,878 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10771.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:01:31,713 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10772.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:01:40,307 INFO [train.py:901] (1/4) Epoch 2, batch 2700, loss[loss=0.3782, simple_loss=0.4088, pruned_loss=0.1738, over 8617.00 frames. ], tot_loss[loss=0.384, simple_loss=0.4146, pruned_loss=0.1767, over 1619989.17 frames. ], batch size: 34, lr: 3.18e-02, grad_scale: 8.0 +2023-02-05 20:01:46,658 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10792.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:01:48,801 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10795.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 20:01:50,102 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10797.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:02:01,031 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.290e+02 4.005e+02 5.458e+02 7.000e+02 2.619e+03, threshold=1.092e+03, percent-clipped=7.0 +2023-02-05 20:02:03,279 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10816.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:02:06,059 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10820.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:02:15,182 INFO [train.py:901] (1/4) Epoch 2, batch 2750, loss[loss=0.3701, simple_loss=0.372, pruned_loss=0.1841, over 7429.00 frames. ], tot_loss[loss=0.3828, simple_loss=0.4142, pruned_loss=0.1757, over 1620489.35 frames. ], batch size: 17, lr: 3.17e-02, grad_scale: 8.0 +2023-02-05 20:02:49,765 INFO [train.py:901] (1/4) Epoch 2, batch 2800, loss[loss=0.3661, simple_loss=0.3929, pruned_loss=0.1697, over 7662.00 frames. ], tot_loss[loss=0.382, simple_loss=0.4129, pruned_loss=0.1755, over 1615661.67 frames. ], batch size: 19, lr: 3.17e-02, grad_scale: 8.0 +2023-02-05 20:02:49,989 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4389, 1.5377, 2.8405, 0.9955, 1.9164, 1.8998, 1.2848, 1.7027], + device='cuda:1'), covar=tensor([0.1276, 0.1401, 0.0350, 0.1960, 0.0954, 0.1433, 0.1372, 0.1146], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0341, 0.0365, 0.0392, 0.0440, 0.0409, 0.0358, 0.0434], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 20:02:51,262 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10886.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:03:03,334 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10903.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:03:10,630 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.367e+02 3.535e+02 4.531e+02 6.001e+02 1.335e+03, threshold=9.062e+02, percent-clipped=2.0 +2023-02-05 20:03:23,141 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10931.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:03:25,025 INFO [train.py:901] (1/4) Epoch 2, batch 2850, loss[loss=0.318, simple_loss=0.3733, pruned_loss=0.1314, over 7660.00 frames. ], tot_loss[loss=0.3815, simple_loss=0.4127, pruned_loss=0.1752, over 1611487.46 frames. ], batch size: 19, lr: 3.16e-02, grad_scale: 8.0 +2023-02-05 20:03:59,104 INFO [train.py:901] (1/4) Epoch 2, batch 2900, loss[loss=0.3203, simple_loss=0.3821, pruned_loss=0.1292, over 8024.00 frames. ], tot_loss[loss=0.3826, simple_loss=0.4138, pruned_loss=0.1757, over 1611721.56 frames. ], batch size: 22, lr: 3.16e-02, grad_scale: 8.0 +2023-02-05 20:04:19,453 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.551e+02 4.216e+02 5.196e+02 6.845e+02 2.226e+03, threshold=1.039e+03, percent-clipped=10.0 +2023-02-05 20:04:34,447 INFO [train.py:901] (1/4) Epoch 2, batch 2950, loss[loss=0.482, simple_loss=0.4776, pruned_loss=0.2432, over 7642.00 frames. ], tot_loss[loss=0.3813, simple_loss=0.4126, pruned_loss=0.1751, over 1608074.97 frames. ], batch size: 19, lr: 3.15e-02, grad_scale: 8.0 +2023-02-05 20:04:39,272 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-05 20:05:08,646 INFO [train.py:901] (1/4) Epoch 2, batch 3000, loss[loss=0.3714, simple_loss=0.4199, pruned_loss=0.1614, over 8253.00 frames. ], tot_loss[loss=0.3815, simple_loss=0.413, pruned_loss=0.175, over 1611790.10 frames. ], batch size: 24, lr: 3.14e-02, grad_scale: 8.0 +2023-02-05 20:05:08,646 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 20:05:24,854 INFO [train.py:935] (1/4) Epoch 2, validation: loss=0.2878, simple_loss=0.369, pruned_loss=0.1033, over 944034.00 frames. +2023-02-05 20:05:24,855 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6470MB +2023-02-05 20:05:40,476 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11106.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:05:45,161 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.542e+02 3.795e+02 4.955e+02 6.193e+02 1.384e+03, threshold=9.910e+02, percent-clipped=4.0 +2023-02-05 20:06:00,081 INFO [train.py:901] (1/4) Epoch 2, batch 3050, loss[loss=0.3554, simple_loss=0.4032, pruned_loss=0.1538, over 8193.00 frames. ], tot_loss[loss=0.3807, simple_loss=0.4129, pruned_loss=0.1743, over 1613056.93 frames. ], batch size: 23, lr: 3.14e-02, grad_scale: 8.0 +2023-02-05 20:06:01,581 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11136.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:03,723 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6771, 1.5525, 1.2970, 1.1072, 1.5682, 1.4027, 1.4633, 1.5728], + device='cuda:1'), covar=tensor([0.1215, 0.1633, 0.2496, 0.2207, 0.0964, 0.2031, 0.1240, 0.1000], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0261, 0.0284, 0.0257, 0.0231, 0.0248, 0.0227, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-02-05 20:06:05,912 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11142.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:14,864 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0364, 2.4476, 3.9268, 3.9098, 3.2524, 2.4230, 1.6787, 2.4196], + device='cuda:1'), covar=tensor([0.0690, 0.0868, 0.0148, 0.0200, 0.0315, 0.0339, 0.0534, 0.0707], + device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0382, 0.0270, 0.0310, 0.0399, 0.0343, 0.0367, 0.0411], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:06:24,331 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11167.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:35,123 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-02-05 20:06:35,407 INFO [train.py:901] (1/4) Epoch 2, batch 3100, loss[loss=0.3853, simple_loss=0.422, pruned_loss=0.1743, over 8332.00 frames. ], tot_loss[loss=0.3781, simple_loss=0.4112, pruned_loss=0.1725, over 1613749.24 frames. ], batch size: 25, lr: 3.13e-02, grad_scale: 8.0 +2023-02-05 20:06:37,609 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11187.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:40,852 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11192.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:55,409 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11212.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:55,869 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.637e+02 3.930e+02 4.987e+02 6.652e+02 1.229e+03, threshold=9.974e+02, percent-clipped=5.0 +2023-02-05 20:07:01,738 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11221.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:07:10,348 INFO [train.py:901] (1/4) Epoch 2, batch 3150, loss[loss=0.4007, simple_loss=0.4444, pruned_loss=0.1785, over 8504.00 frames. ], tot_loss[loss=0.3773, simple_loss=0.4107, pruned_loss=0.1719, over 1615412.95 frames. ], batch size: 26, lr: 3.13e-02, grad_scale: 8.0 +2023-02-05 20:07:20,139 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11247.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 20:07:22,984 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11251.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:07:29,924 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2463, 1.4618, 1.6945, 1.0407, 0.9340, 1.7071, 0.1620, 0.9974], + device='cuda:1'), covar=tensor([0.0873, 0.0461, 0.0381, 0.0643, 0.0894, 0.0311, 0.1714, 0.0814], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0094, 0.0086, 0.0131, 0.0120, 0.0080, 0.0157, 0.0127], + device='cuda:1'), out_proj_covar=tensor([1.1081e-04, 9.6578e-05, 8.2430e-05, 1.2308e-04, 1.2024e-04, 7.6526e-05, + 1.4727e-04, 1.2461e-04], device='cuda:1') +2023-02-05 20:07:34,696 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8889, 1.1072, 4.1046, 1.7341, 3.3113, 3.3698, 3.5104, 3.5551], + device='cuda:1'), covar=tensor([0.0596, 0.3476, 0.0282, 0.1507, 0.1092, 0.0408, 0.0392, 0.0481], + device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0357, 0.0220, 0.0253, 0.0296, 0.0234, 0.0216, 0.0245], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 20:07:46,078 INFO [train.py:901] (1/4) Epoch 2, batch 3200, loss[loss=0.3347, simple_loss=0.3933, pruned_loss=0.1381, over 8145.00 frames. ], tot_loss[loss=0.3767, simple_loss=0.4103, pruned_loss=0.1716, over 1611940.16 frames. ], batch size: 22, lr: 3.12e-02, grad_scale: 8.0 +2023-02-05 20:08:06,208 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 3.889e+02 4.508e+02 6.050e+02 1.565e+03, threshold=9.016e+02, percent-clipped=4.0 +2023-02-05 20:08:21,225 INFO [train.py:901] (1/4) Epoch 2, batch 3250, loss[loss=0.3461, simple_loss=0.3955, pruned_loss=0.1483, over 8355.00 frames. ], tot_loss[loss=0.3723, simple_loss=0.4065, pruned_loss=0.169, over 1610624.29 frames. ], batch size: 24, lr: 3.11e-02, grad_scale: 8.0 +2023-02-05 20:08:39,999 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11362.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 20:08:55,031 INFO [train.py:901] (1/4) Epoch 2, batch 3300, loss[loss=0.3435, simple_loss=0.3825, pruned_loss=0.1523, over 8099.00 frames. ], tot_loss[loss=0.3738, simple_loss=0.4079, pruned_loss=0.1698, over 1609443.09 frames. ], batch size: 21, lr: 3.11e-02, grad_scale: 8.0 +2023-02-05 20:09:16,012 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.686e+02 3.650e+02 4.417e+02 5.589e+02 1.513e+03, threshold=8.834e+02, percent-clipped=8.0 +2023-02-05 20:09:30,178 INFO [train.py:901] (1/4) Epoch 2, batch 3350, loss[loss=0.3913, simple_loss=0.426, pruned_loss=0.1783, over 8286.00 frames. ], tot_loss[loss=0.3761, simple_loss=0.4097, pruned_loss=0.1712, over 1606913.61 frames. ], batch size: 23, lr: 3.10e-02, grad_scale: 8.0 +2023-02-05 20:09:42,542 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-02-05 20:10:00,577 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11477.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:05,250 INFO [train.py:901] (1/4) Epoch 2, batch 3400, loss[loss=0.3887, simple_loss=0.4258, pruned_loss=0.1758, over 8465.00 frames. ], tot_loss[loss=0.3772, simple_loss=0.4105, pruned_loss=0.172, over 1607259.74 frames. ], batch size: 25, lr: 3.10e-02, grad_scale: 8.0 +2023-02-05 20:10:09,504 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3276, 1.3424, 4.5005, 1.9830, 3.8509, 3.6776, 3.9355, 3.9631], + device='cuda:1'), covar=tensor([0.0280, 0.2717, 0.0199, 0.1139, 0.0759, 0.0281, 0.0250, 0.0330], + device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0365, 0.0222, 0.0252, 0.0310, 0.0244, 0.0225, 0.0256], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 20:10:17,703 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11502.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:21,207 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11507.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:25,780 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.120e+02 3.730e+02 4.591e+02 5.662e+02 1.223e+03, threshold=9.181e+02, percent-clipped=5.0 +2023-02-05 20:10:39,494 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:40,653 INFO [train.py:901] (1/4) Epoch 2, batch 3450, loss[loss=0.37, simple_loss=0.4027, pruned_loss=0.1687, over 7936.00 frames. ], tot_loss[loss=0.3775, simple_loss=0.411, pruned_loss=0.172, over 1610064.95 frames. ], batch size: 20, lr: 3.09e-02, grad_scale: 8.0 +2023-02-05 20:10:42,063 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11536.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:51,626 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11550.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:11:15,658 INFO [train.py:901] (1/4) Epoch 2, batch 3500, loss[loss=0.3235, simple_loss=0.359, pruned_loss=0.144, over 7530.00 frames. ], tot_loss[loss=0.3768, simple_loss=0.4103, pruned_loss=0.1717, over 1605115.82 frames. ], batch size: 18, lr: 3.09e-02, grad_scale: 8.0 +2023-02-05 20:11:19,136 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.6695, 0.9980, 3.7487, 1.3194, 3.1667, 3.0746, 3.2458, 3.2249], + device='cuda:1'), covar=tensor([0.0325, 0.3205, 0.0247, 0.1426, 0.0919, 0.0378, 0.0344, 0.0466], + device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0370, 0.0225, 0.0258, 0.0307, 0.0246, 0.0228, 0.0255], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 20:11:35,939 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.906e+02 4.071e+02 4.877e+02 6.297e+02 1.257e+03, threshold=9.753e+02, percent-clipped=3.0 +2023-02-05 20:11:39,549 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11618.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 20:11:40,718 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-05 20:11:50,751 INFO [train.py:901] (1/4) Epoch 2, batch 3550, loss[loss=0.4031, simple_loss=0.436, pruned_loss=0.1851, over 8445.00 frames. ], tot_loss[loss=0.3753, simple_loss=0.4085, pruned_loss=0.171, over 1603298.48 frames. ], batch size: 27, lr: 3.08e-02, grad_scale: 8.0 +2023-02-05 20:11:57,567 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11643.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:12:02,967 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11651.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:12:04,328 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11653.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:12:25,617 INFO [train.py:901] (1/4) Epoch 2, batch 3600, loss[loss=0.3924, simple_loss=0.4295, pruned_loss=0.1776, over 8021.00 frames. ], tot_loss[loss=0.3732, simple_loss=0.4074, pruned_loss=0.1694, over 1606874.21 frames. ], batch size: 22, lr: 3.08e-02, grad_scale: 8.0 +2023-02-05 20:12:30,444 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7013, 1.2264, 2.9991, 1.3926, 2.0252, 3.4023, 3.1498, 2.9718], + device='cuda:1'), covar=tensor([0.1277, 0.1768, 0.0353, 0.1809, 0.0760, 0.0194, 0.0235, 0.0466], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0267, 0.0177, 0.0252, 0.0202, 0.0150, 0.0146, 0.0217], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:12:39,209 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-05 20:12:45,382 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.676e+02 3.688e+02 4.691e+02 6.662e+02 1.491e+03, threshold=9.383e+02, percent-clipped=3.0 +2023-02-05 20:12:48,295 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11717.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:12:59,434 INFO [train.py:901] (1/4) Epoch 2, batch 3650, loss[loss=0.347, simple_loss=0.3794, pruned_loss=0.1573, over 8357.00 frames. ], tot_loss[loss=0.3731, simple_loss=0.4072, pruned_loss=0.1695, over 1603039.71 frames. ], batch size: 24, lr: 3.07e-02, grad_scale: 8.0 +2023-02-05 20:13:04,729 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3858, 4.5738, 3.8823, 1.8727, 3.7010, 3.6231, 4.1195, 3.2437], + device='cuda:1'), covar=tensor([0.0951, 0.0452, 0.0896, 0.4068, 0.0611, 0.0586, 0.1121, 0.0755], + device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0206, 0.0246, 0.0331, 0.0221, 0.0170, 0.0236, 0.0161], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 20:13:23,237 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.4633, 5.6617, 4.9275, 1.7961, 4.8667, 4.8292, 5.1436, 4.5928], + device='cuda:1'), covar=tensor([0.0806, 0.0319, 0.0723, 0.4292, 0.0475, 0.0509, 0.0880, 0.0421], + device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0209, 0.0249, 0.0336, 0.0224, 0.0171, 0.0238, 0.0161], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 20:13:23,559 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-02-05 20:13:33,823 INFO [train.py:901] (1/4) Epoch 2, batch 3700, loss[loss=0.3364, simple_loss=0.3714, pruned_loss=0.1508, over 7706.00 frames. ], tot_loss[loss=0.3755, simple_loss=0.4094, pruned_loss=0.1708, over 1610728.03 frames. ], batch size: 18, lr: 3.06e-02, grad_scale: 8.0 +2023-02-05 20:13:44,414 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-05 20:13:46,973 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.33 vs. limit=5.0 +2023-02-05 20:13:53,753 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.910e+02 4.224e+02 5.211e+02 6.213e+02 2.304e+03, threshold=1.042e+03, percent-clipped=10.0 +2023-02-05 20:14:08,511 INFO [train.py:901] (1/4) Epoch 2, batch 3750, loss[loss=0.38, simple_loss=0.4106, pruned_loss=0.1747, over 8233.00 frames. ], tot_loss[loss=0.3744, simple_loss=0.4085, pruned_loss=0.1702, over 1611190.79 frames. ], batch size: 22, lr: 3.06e-02, grad_scale: 8.0 +2023-02-05 20:14:08,621 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11834.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:14:28,576 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11864.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:14:43,027 INFO [train.py:901] (1/4) Epoch 2, batch 3800, loss[loss=0.338, simple_loss=0.3918, pruned_loss=0.1422, over 8454.00 frames. ], tot_loss[loss=0.374, simple_loss=0.4078, pruned_loss=0.1701, over 1607967.39 frames. ], batch size: 27, lr: 3.05e-02, grad_scale: 8.0 +2023-02-05 20:14:49,683 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11894.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:14:58,770 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11907.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:15:02,633 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.750e+02 4.056e+02 4.773e+02 6.198e+02 1.391e+03, threshold=9.546e+02, percent-clipped=3.0 +2023-02-05 20:15:07,557 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9423, 2.3406, 2.0798, 2.7962, 1.3081, 1.2252, 1.7687, 2.1940], + device='cuda:1'), covar=tensor([0.1232, 0.1146, 0.1235, 0.0430, 0.2074, 0.2118, 0.2026, 0.1154], + device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0337, 0.0327, 0.0214, 0.0330, 0.0329, 0.0372, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0005, 0.0004], + device='cuda:1') +2023-02-05 20:15:16,348 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11932.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:15:17,489 INFO [train.py:901] (1/4) Epoch 2, batch 3850, loss[loss=0.4148, simple_loss=0.4517, pruned_loss=0.1889, over 8033.00 frames. ], tot_loss[loss=0.3743, simple_loss=0.4087, pruned_loss=0.17, over 1610717.01 frames. ], batch size: 22, lr: 3.05e-02, grad_scale: 8.0 +2023-02-05 20:15:20,311 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11938.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:15:39,746 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11966.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:15:47,063 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-05 20:15:51,653 INFO [train.py:901] (1/4) Epoch 2, batch 3900, loss[loss=0.369, simple_loss=0.4138, pruned_loss=0.1621, over 8619.00 frames. ], tot_loss[loss=0.3745, simple_loss=0.4092, pruned_loss=0.1699, over 1610223.45 frames. ], batch size: 31, lr: 3.04e-02, grad_scale: 8.0 +2023-02-05 20:16:01,111 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11997.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:16:06,092 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12002.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:16:10,874 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:16:13,175 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.113e+02 3.926e+02 4.686e+02 5.678e+02 1.222e+03, threshold=9.373e+02, percent-clipped=4.0 +2023-02-05 20:16:28,134 INFO [train.py:901] (1/4) Epoch 2, batch 3950, loss[loss=0.364, simple_loss=0.3761, pruned_loss=0.1759, over 7432.00 frames. ], tot_loss[loss=0.3732, simple_loss=0.4082, pruned_loss=0.169, over 1612125.04 frames. ], batch size: 17, lr: 3.04e-02, grad_scale: 8.0 +2023-02-05 20:16:46,992 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12061.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:17:02,485 INFO [train.py:901] (1/4) Epoch 2, batch 4000, loss[loss=0.3556, simple_loss=0.386, pruned_loss=0.1626, over 7195.00 frames. ], tot_loss[loss=0.3721, simple_loss=0.4079, pruned_loss=0.1681, over 1615536.86 frames. ], batch size: 16, lr: 3.03e-02, grad_scale: 8.0 +2023-02-05 20:17:09,214 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12094.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:17:22,644 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12112.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:17:23,112 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.955e+02 4.453e+02 5.904e+02 7.845e+02 2.502e+03, threshold=1.181e+03, percent-clipped=13.0 +2023-02-05 20:17:36,860 INFO [train.py:901] (1/4) Epoch 2, batch 4050, loss[loss=0.3433, simple_loss=0.3925, pruned_loss=0.1471, over 8286.00 frames. ], tot_loss[loss=0.3748, simple_loss=0.41, pruned_loss=0.1698, over 1616158.73 frames. ], batch size: 23, lr: 3.03e-02, grad_scale: 16.0 +2023-02-05 20:18:06,024 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12176.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:18:06,916 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 +2023-02-05 20:18:07,288 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12178.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:18:11,164 INFO [train.py:901] (1/4) Epoch 2, batch 4100, loss[loss=0.362, simple_loss=0.4027, pruned_loss=0.1607, over 7796.00 frames. ], tot_loss[loss=0.3761, simple_loss=0.4105, pruned_loss=0.1709, over 1614388.62 frames. ], batch size: 19, lr: 3.02e-02, grad_scale: 16.0 +2023-02-05 20:18:16,106 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9664, 1.5300, 1.3158, 1.2465, 1.9251, 1.4840, 1.3621, 1.6402], + device='cuda:1'), covar=tensor([0.0898, 0.1572, 0.2082, 0.1692, 0.0739, 0.1692, 0.1080, 0.0792], + device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0254, 0.0275, 0.0246, 0.0223, 0.0243, 0.0216, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], + device='cuda:1') +2023-02-05 20:18:16,882 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8083, 1.9181, 2.2676, 0.8618, 2.1515, 1.5982, 0.4982, 1.9286], + device='cuda:1'), covar=tensor([0.0102, 0.0072, 0.0073, 0.0172, 0.0088, 0.0241, 0.0277, 0.0081], + device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0125, 0.0121, 0.0183, 0.0129, 0.0233, 0.0188, 0.0160], + device='cuda:1'), out_proj_covar=tensor([1.1337e-04, 7.8511e-05, 8.0131e-05, 1.1457e-04, 8.6559e-05, 1.5844e-04, + 1.2213e-04, 1.0361e-04], device='cuda:1') +2023-02-05 20:18:27,597 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12208.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:18:28,926 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6768, 1.3025, 3.3536, 1.2449, 2.2110, 3.8339, 3.4509, 3.2379], + device='cuda:1'), covar=tensor([0.1166, 0.1711, 0.0316, 0.1991, 0.0791, 0.0220, 0.0282, 0.0439], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0255, 0.0177, 0.0249, 0.0191, 0.0154, 0.0143, 0.0215], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:18:30,897 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.458e+02 3.728e+02 4.672e+02 5.863e+02 2.072e+03, threshold=9.344e+02, percent-clipped=1.0 +2023-02-05 20:18:47,032 INFO [train.py:901] (1/4) Epoch 2, batch 4150, loss[loss=0.3691, simple_loss=0.4281, pruned_loss=0.155, over 8362.00 frames. ], tot_loss[loss=0.3757, simple_loss=0.4109, pruned_loss=0.1702, over 1618956.19 frames. ], batch size: 24, lr: 3.02e-02, grad_scale: 16.0 +2023-02-05 20:19:08,905 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12265.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:20,418 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12282.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:21,719 INFO [train.py:901] (1/4) Epoch 2, batch 4200, loss[loss=0.3195, simple_loss=0.3887, pruned_loss=0.1251, over 8280.00 frames. ], tot_loss[loss=0.3749, simple_loss=0.4101, pruned_loss=0.1698, over 1619263.62 frames. ], batch size: 23, lr: 3.01e-02, grad_scale: 16.0 +2023-02-05 20:19:25,931 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12290.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:28,632 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12293.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:40,064 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12310.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:42,059 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.324e+02 3.573e+02 4.694e+02 5.833e+02 1.413e+03, threshold=9.388e+02, percent-clipped=6.0 +2023-02-05 20:19:43,535 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-05 20:19:49,252 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12323.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:57,053 INFO [train.py:901] (1/4) Epoch 2, batch 4250, loss[loss=0.3856, simple_loss=0.4255, pruned_loss=0.1729, over 8492.00 frames. ], tot_loss[loss=0.3745, simple_loss=0.4098, pruned_loss=0.1695, over 1616661.25 frames. ], batch size: 26, lr: 3.01e-02, grad_scale: 16.0 +2023-02-05 20:20:06,019 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12346.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:20:06,628 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-05 20:20:20,970 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12368.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:20:32,224 INFO [train.py:901] (1/4) Epoch 2, batch 4300, loss[loss=0.4376, simple_loss=0.4548, pruned_loss=0.2102, over 8640.00 frames. ], tot_loss[loss=0.3737, simple_loss=0.4096, pruned_loss=0.169, over 1616164.96 frames. ], batch size: 39, lr: 3.00e-02, grad_scale: 16.0 +2023-02-05 20:20:38,552 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12393.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:20:41,196 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12397.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:20:53,215 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.244e+02 3.864e+02 4.648e+02 5.983e+02 1.525e+03, threshold=9.296e+02, percent-clipped=6.0 +2023-02-05 20:21:00,853 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12425.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:05,798 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:06,902 INFO [train.py:901] (1/4) Epoch 2, batch 4350, loss[loss=0.3822, simple_loss=0.4067, pruned_loss=0.1789, over 8656.00 frames. ], tot_loss[loss=0.3715, simple_loss=0.4075, pruned_loss=0.1677, over 1611971.84 frames. ], batch size: 49, lr: 2.99e-02, grad_scale: 8.0 +2023-02-05 20:21:09,316 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-02-05 20:21:09,737 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12438.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:23,515 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12457.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:26,041 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12461.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:28,052 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12464.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:38,137 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-05 20:21:42,134 INFO [train.py:901] (1/4) Epoch 2, batch 4400, loss[loss=0.4562, simple_loss=0.4507, pruned_loss=0.2308, over 7941.00 frames. ], tot_loss[loss=0.3714, simple_loss=0.407, pruned_loss=0.1679, over 1612056.26 frames. ], batch size: 20, lr: 2.99e-02, grad_scale: 8.0 +2023-02-05 20:22:02,397 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.494e+02 4.041e+02 4.964e+02 6.742e+02 1.213e+03, threshold=9.928e+02, percent-clipped=4.0 +2023-02-05 20:22:16,723 INFO [train.py:901] (1/4) Epoch 2, batch 4450, loss[loss=0.3501, simple_loss=0.3814, pruned_loss=0.1594, over 8079.00 frames. ], tot_loss[loss=0.3713, simple_loss=0.407, pruned_loss=0.1679, over 1611699.80 frames. ], batch size: 21, lr: 2.98e-02, grad_scale: 8.0 +2023-02-05 20:22:17,404 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-05 20:22:27,255 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12549.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:22:28,829 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-02-05 20:22:29,912 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12553.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:22:45,079 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12574.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:22:49,126 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12579.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:22:52,408 INFO [train.py:901] (1/4) Epoch 2, batch 4500, loss[loss=0.4176, simple_loss=0.4315, pruned_loss=0.2018, over 8495.00 frames. ], tot_loss[loss=0.3694, simple_loss=0.4059, pruned_loss=0.1665, over 1613611.31 frames. ], batch size: 26, lr: 2.98e-02, grad_scale: 8.0 +2023-02-05 20:23:06,028 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:23:12,547 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-05 20:23:13,222 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.312e+02 4.309e+02 5.092e+02 6.256e+02 1.421e+03, threshold=1.018e+03, percent-clipped=5.0 +2023-02-05 20:23:25,280 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5434, 2.0502, 3.5312, 3.0474, 2.6059, 2.0706, 1.4379, 1.9047], + device='cuda:1'), covar=tensor([0.0597, 0.0736, 0.0118, 0.0227, 0.0317, 0.0305, 0.0442, 0.0608], + device='cuda:1'), in_proj_covar=tensor([0.0492, 0.0402, 0.0293, 0.0337, 0.0420, 0.0363, 0.0385, 0.0419], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:23:27,083 INFO [train.py:901] (1/4) Epoch 2, batch 4550, loss[loss=0.3917, simple_loss=0.4252, pruned_loss=0.179, over 8504.00 frames. ], tot_loss[loss=0.3683, simple_loss=0.4044, pruned_loss=0.1661, over 1610294.99 frames. ], batch size: 26, lr: 2.97e-02, grad_scale: 8.0 +2023-02-05 20:23:40,670 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12653.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:23:57,691 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12678.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:23:59,767 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12681.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:24:01,607 INFO [train.py:901] (1/4) Epoch 2, batch 4600, loss[loss=0.3519, simple_loss=0.4062, pruned_loss=0.1488, over 8521.00 frames. ], tot_loss[loss=0.3679, simple_loss=0.4042, pruned_loss=0.1658, over 1610315.84 frames. ], batch size: 26, lr: 2.97e-02, grad_scale: 8.0 +2023-02-05 20:24:17,918 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12706.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:24:23,143 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.401e+02 3.817e+02 4.647e+02 5.826e+02 1.354e+03, threshold=9.293e+02, percent-clipped=3.0 +2023-02-05 20:24:25,437 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12717.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:24:26,985 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.13 vs. limit=5.0 +2023-02-05 20:24:31,992 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.5874, 2.5804, 4.1258, 4.3168, 3.3072, 2.5892, 1.8100, 2.2889], + device='cuda:1'), covar=tensor([0.0471, 0.0774, 0.0115, 0.0179, 0.0263, 0.0257, 0.0423, 0.0565], + device='cuda:1'), in_proj_covar=tensor([0.0485, 0.0402, 0.0291, 0.0341, 0.0419, 0.0362, 0.0383, 0.0417], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:24:37,073 INFO [train.py:901] (1/4) Epoch 2, batch 4650, loss[loss=0.3392, simple_loss=0.381, pruned_loss=0.1487, over 7803.00 frames. ], tot_loss[loss=0.3681, simple_loss=0.404, pruned_loss=0.1661, over 1609395.34 frames. ], batch size: 20, lr: 2.96e-02, grad_scale: 8.0 +2023-02-05 20:24:42,619 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12742.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:24:46,544 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5493, 1.1529, 4.5342, 1.9326, 4.0414, 3.7942, 3.9922, 3.9338], + device='cuda:1'), covar=tensor([0.0267, 0.3158, 0.0219, 0.1398, 0.0758, 0.0338, 0.0313, 0.0361], + device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0376, 0.0234, 0.0267, 0.0322, 0.0259, 0.0243, 0.0264], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 20:24:47,918 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4411, 1.7668, 1.3521, 1.1730, 1.9954, 1.4739, 1.6492, 1.5806], + device='cuda:1'), covar=tensor([0.0899, 0.1554, 0.2190, 0.1860, 0.0788, 0.1797, 0.1098, 0.0916], + device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0253, 0.0278, 0.0248, 0.0224, 0.0245, 0.0216, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004], + device='cuda:1') +2023-02-05 20:25:09,730 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12781.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:25:11,634 INFO [train.py:901] (1/4) Epoch 2, batch 4700, loss[loss=0.3282, simple_loss=0.3727, pruned_loss=0.1419, over 8240.00 frames. ], tot_loss[loss=0.365, simple_loss=0.4017, pruned_loss=0.1641, over 1611444.45 frames. ], batch size: 22, lr: 2.96e-02, grad_scale: 8.0 +2023-02-05 20:25:28,058 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12808.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:25:29,556 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12809.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:25:32,791 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.540e+02 4.122e+02 5.358e+02 6.927e+02 1.344e+03, threshold=1.072e+03, percent-clipped=8.0 +2023-02-05 20:25:47,171 INFO [train.py:901] (1/4) Epoch 2, batch 4750, loss[loss=0.4306, simple_loss=0.4365, pruned_loss=0.2124, over 7325.00 frames. ], tot_loss[loss=0.3662, simple_loss=0.4021, pruned_loss=0.1651, over 1607124.37 frames. ], batch size: 71, lr: 2.95e-02, grad_scale: 8.0 +2023-02-05 20:25:47,388 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12834.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:26:01,688 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.9373, 4.1416, 3.6295, 1.8091, 3.5501, 3.5054, 3.7212, 2.9867], + device='cuda:1'), covar=tensor([0.0817, 0.0428, 0.0739, 0.3556, 0.0497, 0.0568, 0.0852, 0.0592], + device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0209, 0.0253, 0.0333, 0.0230, 0.0173, 0.0235, 0.0156], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 20:26:12,046 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4601, 2.1293, 3.6776, 1.0865, 2.6406, 1.7528, 1.5387, 2.1674], + device='cuda:1'), covar=tensor([0.0974, 0.0974, 0.0301, 0.1521, 0.0903, 0.1468, 0.0883, 0.1256], + device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0365, 0.0403, 0.0423, 0.0481, 0.0425, 0.0379, 0.0479], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 20:26:18,679 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-05 20:26:20,730 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-05 20:26:22,726 INFO [train.py:901] (1/4) Epoch 2, batch 4800, loss[loss=0.3423, simple_loss=0.3741, pruned_loss=0.1553, over 7806.00 frames. ], tot_loss[loss=0.3665, simple_loss=0.4023, pruned_loss=0.1654, over 1603996.99 frames. ], batch size: 20, lr: 2.95e-02, grad_scale: 8.0 +2023-02-05 20:26:25,597 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4425, 4.6585, 4.0177, 1.5816, 3.9251, 4.0892, 4.1539, 3.4725], + device='cuda:1'), covar=tensor([0.0871, 0.0296, 0.0659, 0.4239, 0.0477, 0.0453, 0.0821, 0.0485], + device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0212, 0.0259, 0.0344, 0.0236, 0.0180, 0.0243, 0.0161], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 20:26:36,806 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7794, 1.2492, 3.9645, 1.6947, 3.3637, 3.2200, 3.4574, 3.4917], + device='cuda:1'), covar=tensor([0.0370, 0.3329, 0.0226, 0.1635, 0.0929, 0.0403, 0.0387, 0.0431], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0387, 0.0238, 0.0275, 0.0331, 0.0264, 0.0249, 0.0268], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 20:26:43,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.314e+02 3.678e+02 4.471e+02 5.888e+02 1.234e+03, threshold=8.941e+02, percent-clipped=3.0 +2023-02-05 20:26:49,701 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12923.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:26:57,686 INFO [train.py:901] (1/4) Epoch 2, batch 4850, loss[loss=0.3601, simple_loss=0.377, pruned_loss=0.1716, over 7690.00 frames. ], tot_loss[loss=0.3654, simple_loss=0.4017, pruned_loss=0.1646, over 1604016.51 frames. ], batch size: 18, lr: 2.94e-02, grad_scale: 8.0 +2023-02-05 20:27:05,034 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0941, 1.3928, 1.5358, 1.0288, 0.8107, 1.5353, 0.2649, 0.7764], + device='cuda:1'), covar=tensor([0.1156, 0.0837, 0.0604, 0.1035, 0.1140, 0.0526, 0.2338, 0.1228], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0095, 0.0087, 0.0149, 0.0125, 0.0086, 0.0159, 0.0125], + device='cuda:1'), out_proj_covar=tensor([1.1890e-04, 1.0350e-04, 9.0018e-05, 1.4585e-04, 1.3103e-04, 8.8109e-05, + 1.5848e-04, 1.3131e-04], device='cuda:1') +2023-02-05 20:27:12,750 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-05 20:27:32,108 INFO [train.py:901] (1/4) Epoch 2, batch 4900, loss[loss=0.3173, simple_loss=0.3431, pruned_loss=0.1458, over 7799.00 frames. ], tot_loss[loss=0.3668, simple_loss=0.4027, pruned_loss=0.1654, over 1610669.41 frames. ], batch size: 19, lr: 2.94e-02, grad_scale: 8.0 +2023-02-05 20:27:43,685 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-02-05 20:27:53,285 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 4.170e+02 5.532e+02 7.452e+02 1.588e+03, threshold=1.106e+03, percent-clipped=9.0 +2023-02-05 20:28:06,716 INFO [train.py:901] (1/4) Epoch 2, batch 4950, loss[loss=0.3155, simple_loss=0.3661, pruned_loss=0.1324, over 8022.00 frames. ], tot_loss[loss=0.3676, simple_loss=0.4029, pruned_loss=0.1661, over 1607099.53 frames. ], batch size: 22, lr: 2.93e-02, grad_scale: 8.0 +2023-02-05 20:28:10,874 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-05 20:28:22,843 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8901, 1.6479, 2.5559, 0.8744, 2.4519, 1.6603, 1.1490, 1.9868], + device='cuda:1'), covar=tensor([0.0214, 0.0118, 0.0141, 0.0236, 0.0179, 0.0320, 0.0310, 0.0141], + device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0139, 0.0125, 0.0187, 0.0134, 0.0242, 0.0194, 0.0165], + device='cuda:1'), out_proj_covar=tensor([1.1356e-04, 8.4579e-05, 7.9256e-05, 1.1359e-04, 8.6796e-05, 1.5971e-04, + 1.2185e-04, 1.0173e-04], device='cuda:1') +2023-02-05 20:28:41,855 INFO [train.py:901] (1/4) Epoch 2, batch 5000, loss[loss=0.2976, simple_loss=0.3483, pruned_loss=0.1235, over 7974.00 frames. ], tot_loss[loss=0.3688, simple_loss=0.4044, pruned_loss=0.1666, over 1613466.63 frames. ], batch size: 21, lr: 2.93e-02, grad_scale: 8.0 +2023-02-05 20:29:02,469 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.193e+02 4.113e+02 5.050e+02 6.511e+02 1.788e+03, threshold=1.010e+03, percent-clipped=5.0 +2023-02-05 20:29:09,704 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=13125.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:29:15,847 INFO [train.py:901] (1/4) Epoch 2, batch 5050, loss[loss=0.3605, simple_loss=0.4134, pruned_loss=0.1538, over 8108.00 frames. ], tot_loss[loss=0.37, simple_loss=0.4052, pruned_loss=0.1674, over 1612746.61 frames. ], batch size: 23, lr: 2.92e-02, grad_scale: 4.0 +2023-02-05 20:29:21,929 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2385, 1.6019, 3.1925, 0.8259, 2.8490, 1.9872, 0.9719, 1.7290], + device='cuda:1'), covar=tensor([0.0132, 0.0097, 0.0079, 0.0185, 0.0103, 0.0249, 0.0253, 0.0117], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0135, 0.0125, 0.0182, 0.0130, 0.0238, 0.0187, 0.0161], + device='cuda:1'), out_proj_covar=tensor([1.1239e-04, 8.1977e-05, 7.9614e-05, 1.0961e-04, 8.3239e-05, 1.5611e-04, + 1.1633e-04, 9.8533e-05], device='cuda:1') +2023-02-05 20:29:33,601 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-02-05 20:29:47,498 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13179.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:29:47,977 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-05 20:29:50,623 INFO [train.py:901] (1/4) Epoch 2, batch 5100, loss[loss=0.4024, simple_loss=0.4221, pruned_loss=0.1913, over 8631.00 frames. ], tot_loss[loss=0.371, simple_loss=0.4054, pruned_loss=0.1682, over 1611272.25 frames. ], batch size: 34, lr: 2.92e-02, grad_scale: 4.0 +2023-02-05 20:30:04,620 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13204.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:30:09,125 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4582, 1.8440, 2.9436, 1.1596, 2.0877, 1.6140, 1.4280, 1.6801], + device='cuda:1'), covar=tensor([0.0837, 0.0888, 0.0297, 0.1322, 0.0774, 0.1362, 0.0742, 0.1101], + device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0364, 0.0401, 0.0423, 0.0487, 0.0426, 0.0378, 0.0477], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 20:30:11,536 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.243e+02 3.930e+02 4.883e+02 5.892e+02 1.355e+03, threshold=9.766e+02, percent-clipped=3.0 +2023-02-05 20:30:15,632 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1174, 1.6689, 1.2904, 1.2003, 1.6066, 1.3991, 1.5243, 1.6498], + device='cuda:1'), covar=tensor([0.0952, 0.1491, 0.2121, 0.1749, 0.0904, 0.1755, 0.1150, 0.0841], + device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0255, 0.0282, 0.0251, 0.0225, 0.0247, 0.0218, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004], + device='cuda:1') +2023-02-05 20:30:24,603 INFO [train.py:901] (1/4) Epoch 2, batch 5150, loss[loss=0.3481, simple_loss=0.3897, pruned_loss=0.1533, over 8073.00 frames. ], tot_loss[loss=0.3702, simple_loss=0.4051, pruned_loss=0.1676, over 1612537.07 frames. ], batch size: 21, lr: 2.91e-02, grad_scale: 4.0 +2023-02-05 20:30:28,689 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13240.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:30:39,616 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0783, 2.3555, 4.0141, 3.8600, 3.1423, 2.6010, 1.8220, 2.3301], + device='cuda:1'), covar=tensor([0.0638, 0.0823, 0.0177, 0.0241, 0.0336, 0.0304, 0.0448, 0.0699], + device='cuda:1'), in_proj_covar=tensor([0.0498, 0.0417, 0.0311, 0.0349, 0.0440, 0.0387, 0.0407, 0.0435], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:30:59,003 INFO [train.py:901] (1/4) Epoch 2, batch 5200, loss[loss=0.3705, simple_loss=0.3824, pruned_loss=0.1793, over 7429.00 frames. ], tot_loss[loss=0.3696, simple_loss=0.4049, pruned_loss=0.1672, over 1616933.22 frames. ], batch size: 17, lr: 2.91e-02, grad_scale: 8.0 +2023-02-05 20:30:59,193 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4019, 1.4487, 3.9795, 1.8078, 2.3476, 4.7985, 4.1898, 4.1021], + device='cuda:1'), covar=tensor([0.1041, 0.1589, 0.0265, 0.1751, 0.0732, 0.0195, 0.0269, 0.0494], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0257, 0.0177, 0.0251, 0.0189, 0.0152, 0.0145, 0.0218], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:31:06,812 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-02-05 20:31:20,912 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.071e+02 4.339e+02 5.206e+02 6.705e+02 1.063e+03, threshold=1.041e+03, percent-clipped=3.0 +2023-02-05 20:31:33,618 INFO [train.py:901] (1/4) Epoch 2, batch 5250, loss[loss=0.3399, simple_loss=0.3952, pruned_loss=0.1423, over 8334.00 frames. ], tot_loss[loss=0.3698, simple_loss=0.4053, pruned_loss=0.1672, over 1618761.65 frames. ], batch size: 26, lr: 2.91e-02, grad_scale: 8.0 +2023-02-05 20:31:33,949 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-05 20:31:42,984 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-05 20:31:48,409 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2304, 1.1132, 2.2334, 1.0587, 2.1331, 2.4416, 2.1670, 2.0411], + device='cuda:1'), covar=tensor([0.1077, 0.1265, 0.0376, 0.1671, 0.0450, 0.0313, 0.0412, 0.0633], + device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0251, 0.0175, 0.0245, 0.0189, 0.0151, 0.0146, 0.0216], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:32:01,984 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.74 vs. limit=5.0 +2023-02-05 20:32:07,576 INFO [train.py:901] (1/4) Epoch 2, batch 5300, loss[loss=0.3931, simple_loss=0.4307, pruned_loss=0.1778, over 8328.00 frames. ], tot_loss[loss=0.3676, simple_loss=0.404, pruned_loss=0.1657, over 1617893.96 frames. ], batch size: 25, lr: 2.90e-02, grad_scale: 8.0 +2023-02-05 20:32:07,737 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9818, 1.5912, 3.3137, 1.2305, 2.2490, 3.7512, 3.3845, 3.2193], + device='cuda:1'), covar=tensor([0.1075, 0.1271, 0.0282, 0.1890, 0.0611, 0.0199, 0.0225, 0.0493], + device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0252, 0.0176, 0.0249, 0.0189, 0.0152, 0.0145, 0.0218], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:32:29,083 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 3.821e+02 4.884e+02 6.417e+02 1.823e+03, threshold=9.767e+02, percent-clipped=6.0 +2023-02-05 20:32:42,521 INFO [train.py:901] (1/4) Epoch 2, batch 5350, loss[loss=0.3645, simple_loss=0.4079, pruned_loss=0.1605, over 8253.00 frames. ], tot_loss[loss=0.3697, simple_loss=0.4056, pruned_loss=0.1669, over 1619649.71 frames. ], batch size: 24, lr: 2.90e-02, grad_scale: 8.0 +2023-02-05 20:32:46,414 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-02-05 20:33:16,587 INFO [train.py:901] (1/4) Epoch 2, batch 5400, loss[loss=0.3898, simple_loss=0.4262, pruned_loss=0.1767, over 8653.00 frames. ], tot_loss[loss=0.3696, simple_loss=0.4055, pruned_loss=0.1668, over 1617070.63 frames. ], batch size: 34, lr: 2.89e-02, grad_scale: 8.0 +2023-02-05 20:33:24,899 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13496.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:33:38,014 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.355e+02 3.820e+02 4.559e+02 5.766e+02 1.205e+03, threshold=9.119e+02, percent-clipped=6.0 +2023-02-05 20:33:43,037 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13521.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:33:51,332 INFO [train.py:901] (1/4) Epoch 2, batch 5450, loss[loss=0.3668, simple_loss=0.4032, pruned_loss=0.1652, over 8193.00 frames. ], tot_loss[loss=0.3689, simple_loss=0.4054, pruned_loss=0.1662, over 1618340.82 frames. ], batch size: 23, lr: 2.89e-02, grad_scale: 8.0 +2023-02-05 20:34:25,970 INFO [train.py:901] (1/4) Epoch 2, batch 5500, loss[loss=0.3988, simple_loss=0.4331, pruned_loss=0.1822, over 8515.00 frames. ], tot_loss[loss=0.3706, simple_loss=0.4063, pruned_loss=0.1675, over 1621881.09 frames. ], batch size: 26, lr: 2.88e-02, grad_scale: 8.0 +2023-02-05 20:34:28,062 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-05 20:34:46,540 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.267e+02 3.726e+02 4.817e+02 6.308e+02 1.682e+03, threshold=9.635e+02, percent-clipped=6.0 +2023-02-05 20:34:59,984 INFO [train.py:901] (1/4) Epoch 2, batch 5550, loss[loss=0.3184, simple_loss=0.3713, pruned_loss=0.1327, over 7811.00 frames. ], tot_loss[loss=0.3676, simple_loss=0.4038, pruned_loss=0.1657, over 1618475.15 frames. ], batch size: 20, lr: 2.88e-02, grad_scale: 8.0 +2023-02-05 20:35:35,316 INFO [train.py:901] (1/4) Epoch 2, batch 5600, loss[loss=0.3856, simple_loss=0.4234, pruned_loss=0.1738, over 8501.00 frames. ], tot_loss[loss=0.3674, simple_loss=0.4033, pruned_loss=0.1658, over 1613700.58 frames. ], batch size: 26, lr: 2.87e-02, grad_scale: 8.0 +2023-02-05 20:35:55,773 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 3.833e+02 4.619e+02 6.071e+02 1.383e+03, threshold=9.238e+02, percent-clipped=5.0 +2023-02-05 20:36:08,576 INFO [train.py:901] (1/4) Epoch 2, batch 5650, loss[loss=0.3628, simple_loss=0.3967, pruned_loss=0.1644, over 8509.00 frames. ], tot_loss[loss=0.3687, simple_loss=0.4046, pruned_loss=0.1664, over 1614805.40 frames. ], batch size: 49, lr: 2.87e-02, grad_scale: 8.0 +2023-02-05 20:36:23,364 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13755.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:36:34,194 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-05 20:36:43,557 INFO [train.py:901] (1/4) Epoch 2, batch 5700, loss[loss=0.2859, simple_loss=0.3448, pruned_loss=0.1135, over 8132.00 frames. ], tot_loss[loss=0.367, simple_loss=0.4035, pruned_loss=0.1652, over 1614980.35 frames. ], batch size: 22, lr: 2.86e-02, grad_scale: 8.0 +2023-02-05 20:37:05,583 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.337e+02 4.261e+02 5.123e+02 6.631e+02 2.352e+03, threshold=1.025e+03, percent-clipped=5.0 +2023-02-05 20:37:18,900 INFO [train.py:901] (1/4) Epoch 2, batch 5750, loss[loss=0.3299, simple_loss=0.3631, pruned_loss=0.1484, over 7788.00 frames. ], tot_loss[loss=0.3689, simple_loss=0.4052, pruned_loss=0.1663, over 1615160.97 frames. ], batch size: 19, lr: 2.86e-02, grad_scale: 8.0 +2023-02-05 20:37:33,850 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.20 vs. limit=2.0 +2023-02-05 20:37:38,970 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-05 20:37:54,461 INFO [train.py:901] (1/4) Epoch 2, batch 5800, loss[loss=0.4145, simple_loss=0.4468, pruned_loss=0.1912, over 8249.00 frames. ], tot_loss[loss=0.3704, simple_loss=0.4067, pruned_loss=0.167, over 1616220.60 frames. ], batch size: 24, lr: 2.85e-02, grad_scale: 8.0 +2023-02-05 20:38:15,733 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.601e+02 3.784e+02 4.729e+02 6.225e+02 2.390e+03, threshold=9.458e+02, percent-clipped=5.0 +2023-02-05 20:38:29,061 INFO [train.py:901] (1/4) Epoch 2, batch 5850, loss[loss=0.3778, simple_loss=0.4305, pruned_loss=0.1625, over 8241.00 frames. ], tot_loss[loss=0.3694, simple_loss=0.4057, pruned_loss=0.1666, over 1612777.62 frames. ], batch size: 24, lr: 2.85e-02, grad_scale: 8.0 +2023-02-05 20:39:03,909 INFO [train.py:901] (1/4) Epoch 2, batch 5900, loss[loss=0.3501, simple_loss=0.387, pruned_loss=0.1567, over 7803.00 frames. ], tot_loss[loss=0.3672, simple_loss=0.4041, pruned_loss=0.1651, over 1606771.04 frames. ], batch size: 20, lr: 2.84e-02, grad_scale: 8.0 +2023-02-05 20:39:11,156 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9345, 2.2701, 4.6071, 1.1658, 2.7692, 2.1107, 1.6888, 2.0046], + device='cuda:1'), covar=tensor([0.0958, 0.1168, 0.0278, 0.1842, 0.1080, 0.1476, 0.0954, 0.1713], + device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0379, 0.0424, 0.0449, 0.0499, 0.0443, 0.0398, 0.0497], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 20:39:27,081 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.452e+02 3.946e+02 4.724e+02 6.297e+02 1.551e+03, threshold=9.448e+02, percent-clipped=7.0 +2023-02-05 20:39:40,158 INFO [train.py:901] (1/4) Epoch 2, batch 5950, loss[loss=0.3883, simple_loss=0.4268, pruned_loss=0.1749, over 8360.00 frames. ], tot_loss[loss=0.3654, simple_loss=0.403, pruned_loss=0.1639, over 1606408.80 frames. ], batch size: 26, lr: 2.84e-02, grad_scale: 8.0 +2023-02-05 20:39:45,639 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6579, 1.4711, 2.7291, 0.9735, 2.1315, 2.9587, 2.9373, 2.3935], + device='cuda:1'), covar=tensor([0.1399, 0.1805, 0.0525, 0.2655, 0.0774, 0.0506, 0.0349, 0.0976], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0264, 0.0188, 0.0258, 0.0201, 0.0160, 0.0150, 0.0235], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:40:00,196 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=5.54 vs. limit=5.0 +2023-02-05 20:40:07,401 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5484, 3.2514, 2.8040, 3.9896, 1.9937, 1.4853, 2.0304, 3.1576], + device='cuda:1'), covar=tensor([0.1141, 0.1308, 0.1315, 0.0253, 0.1780, 0.2585, 0.2418, 0.1178], + device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0345, 0.0329, 0.0231, 0.0312, 0.0331, 0.0367, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0005, 0.0005, 0.0003, 0.0004, 0.0005, 0.0005, 0.0004], + device='cuda:1') +2023-02-05 20:40:14,643 INFO [train.py:901] (1/4) Epoch 2, batch 6000, loss[loss=0.3535, simple_loss=0.4181, pruned_loss=0.1445, over 8237.00 frames. ], tot_loss[loss=0.365, simple_loss=0.4028, pruned_loss=0.1636, over 1609808.30 frames. ], batch size: 24, lr: 2.84e-02, grad_scale: 8.0 +2023-02-05 20:40:14,643 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 20:40:27,834 INFO [train.py:935] (1/4) Epoch 2, validation: loss=0.2758, simple_loss=0.3606, pruned_loss=0.0955, over 944034.00 frames. +2023-02-05 20:40:27,835 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-05 20:40:31,352 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2913, 1.3604, 2.6580, 0.9427, 2.0150, 2.9541, 2.7274, 2.5270], + device='cuda:1'), covar=tensor([0.1291, 0.1424, 0.0472, 0.2074, 0.0663, 0.0321, 0.0303, 0.0590], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0261, 0.0185, 0.0255, 0.0198, 0.0159, 0.0148, 0.0232], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:40:32,058 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14090.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:40:38,738 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14099.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:40:49,506 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.396e+02 3.733e+02 4.780e+02 6.772e+02 2.203e+03, threshold=9.561e+02, percent-clipped=10.0 +2023-02-05 20:40:53,729 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14121.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:41:02,694 INFO [train.py:901] (1/4) Epoch 2, batch 6050, loss[loss=0.3176, simple_loss=0.3657, pruned_loss=0.1348, over 8079.00 frames. ], tot_loss[loss=0.3638, simple_loss=0.4021, pruned_loss=0.1627, over 1608168.24 frames. ], batch size: 21, lr: 2.83e-02, grad_scale: 8.0 +2023-02-05 20:41:05,430 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14138.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 20:41:08,102 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1173, 1.3294, 1.5815, 1.2615, 0.9238, 1.5084, 0.3152, 0.7166], + device='cuda:1'), covar=tensor([0.2016, 0.1570, 0.0849, 0.1481, 0.2026, 0.0810, 0.3759, 0.1796], + device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0097, 0.0083, 0.0142, 0.0142, 0.0084, 0.0161, 0.0118], + device='cuda:1'), out_proj_covar=tensor([1.1492e-04, 1.1103e-04, 8.9692e-05, 1.4729e-04, 1.5143e-04, 9.2284e-05, + 1.6641e-04, 1.2941e-04], device='cuda:1') +2023-02-05 20:41:12,689 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8515, 2.1143, 2.5950, 1.5092, 1.4998, 2.1853, 0.2953, 1.1360], + device='cuda:1'), covar=tensor([0.1346, 0.1814, 0.0480, 0.1190, 0.1604, 0.0648, 0.3172, 0.1336], + device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0096, 0.0083, 0.0140, 0.0141, 0.0083, 0.0160, 0.0117], + device='cuda:1'), out_proj_covar=tensor([1.1430e-04, 1.1016e-04, 8.9026e-05, 1.4556e-04, 1.5041e-04, 9.1809e-05, + 1.6520e-04, 1.2828e-04], device='cuda:1') +2023-02-05 20:41:34,278 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-02-05 20:41:37,173 INFO [train.py:901] (1/4) Epoch 2, batch 6100, loss[loss=0.3345, simple_loss=0.371, pruned_loss=0.149, over 8076.00 frames. ], tot_loss[loss=0.3646, simple_loss=0.4023, pruned_loss=0.1635, over 1611060.38 frames. ], batch size: 21, lr: 2.83e-02, grad_scale: 8.0 +2023-02-05 20:41:58,419 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14214.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:41:58,918 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.150e+02 3.920e+02 4.920e+02 6.492e+02 2.677e+03, threshold=9.840e+02, percent-clipped=6.0 +2023-02-05 20:42:08,073 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-05 20:42:11,467 INFO [train.py:901] (1/4) Epoch 2, batch 6150, loss[loss=0.4274, simple_loss=0.4477, pruned_loss=0.2036, over 8594.00 frames. ], tot_loss[loss=0.3653, simple_loss=0.4033, pruned_loss=0.1637, over 1613872.33 frames. ], batch size: 34, lr: 2.82e-02, grad_scale: 8.0 +2023-02-05 20:42:46,428 INFO [train.py:901] (1/4) Epoch 2, batch 6200, loss[loss=0.344, simple_loss=0.3722, pruned_loss=0.1578, over 7815.00 frames. ], tot_loss[loss=0.3641, simple_loss=0.4025, pruned_loss=0.1629, over 1618307.27 frames. ], batch size: 20, lr: 2.82e-02, grad_scale: 8.0 +2023-02-05 20:42:58,969 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0238, 2.4101, 3.9170, 3.6046, 2.7877, 2.1902, 1.4224, 2.1544], + device='cuda:1'), covar=tensor([0.0788, 0.1079, 0.0197, 0.0298, 0.0502, 0.0395, 0.0593, 0.0841], + device='cuda:1'), in_proj_covar=tensor([0.0512, 0.0435, 0.0326, 0.0365, 0.0468, 0.0403, 0.0422, 0.0442], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:43:08,134 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.406e+02 3.453e+02 4.846e+02 6.394e+02 2.249e+03, threshold=9.691e+02, percent-clipped=6.0 +2023-02-05 20:43:21,530 INFO [train.py:901] (1/4) Epoch 2, batch 6250, loss[loss=0.4365, simple_loss=0.4606, pruned_loss=0.2062, over 8249.00 frames. ], tot_loss[loss=0.3632, simple_loss=0.4013, pruned_loss=0.1626, over 1616049.10 frames. ], batch size: 24, lr: 2.81e-02, grad_scale: 8.0 +2023-02-05 20:43:55,855 INFO [train.py:901] (1/4) Epoch 2, batch 6300, loss[loss=0.3345, simple_loss=0.3725, pruned_loss=0.1482, over 7703.00 frames. ], tot_loss[loss=0.3655, simple_loss=0.4033, pruned_loss=0.1638, over 1618275.66 frames. ], batch size: 18, lr: 2.81e-02, grad_scale: 8.0 +2023-02-05 20:44:17,506 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.643e+02 3.823e+02 4.655e+02 5.877e+02 1.568e+03, threshold=9.309e+02, percent-clipped=4.0 +2023-02-05 20:44:28,390 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14431.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:44:30,254 INFO [train.py:901] (1/4) Epoch 2, batch 6350, loss[loss=0.3622, simple_loss=0.3829, pruned_loss=0.1708, over 7798.00 frames. ], tot_loss[loss=0.364, simple_loss=0.4014, pruned_loss=0.1633, over 1616473.83 frames. ], batch size: 19, lr: 2.81e-02, grad_scale: 8.0 +2023-02-05 20:44:30,320 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14434.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:44:51,460 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14465.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:44:54,882 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14470.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:45:03,110 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14482.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 20:45:04,270 INFO [train.py:901] (1/4) Epoch 2, batch 6400, loss[loss=0.356, simple_loss=0.4005, pruned_loss=0.1557, over 8103.00 frames. ], tot_loss[loss=0.3656, simple_loss=0.4023, pruned_loss=0.1644, over 1614199.13 frames. ], batch size: 23, lr: 2.80e-02, grad_scale: 8.0 +2023-02-05 20:45:12,400 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14495.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:45:19,137 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14505.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:45:25,558 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.523e+02 3.974e+02 5.065e+02 7.362e+02 1.328e+03, threshold=1.013e+03, percent-clipped=8.0 +2023-02-05 20:45:38,723 INFO [train.py:901] (1/4) Epoch 2, batch 6450, loss[loss=0.3879, simple_loss=0.4204, pruned_loss=0.1777, over 8469.00 frames. ], tot_loss[loss=0.3617, simple_loss=0.3993, pruned_loss=0.162, over 1611870.37 frames. ], batch size: 27, lr: 2.80e-02, grad_scale: 8.0 +2023-02-05 20:45:48,960 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14549.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:46:10,570 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14580.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:46:13,165 INFO [train.py:901] (1/4) Epoch 2, batch 6500, loss[loss=0.389, simple_loss=0.4268, pruned_loss=0.1756, over 8464.00 frames. ], tot_loss[loss=0.3617, simple_loss=0.3991, pruned_loss=0.1621, over 1611976.24 frames. ], batch size: 25, lr: 2.79e-02, grad_scale: 8.0 +2023-02-05 20:46:22,643 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14597.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:46:28,918 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6841, 2.3033, 4.6001, 1.0070, 2.6454, 1.9031, 1.7324, 2.4348], + device='cuda:1'), covar=tensor([0.1257, 0.1436, 0.0366, 0.2281, 0.1305, 0.2050, 0.1087, 0.1715], + device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0382, 0.0434, 0.0461, 0.0506, 0.0457, 0.0405, 0.0508], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 20:46:35,357 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.440e+02 3.999e+02 5.009e+02 6.288e+02 1.522e+03, threshold=1.002e+03, percent-clipped=8.0 +2023-02-05 20:46:48,434 INFO [train.py:901] (1/4) Epoch 2, batch 6550, loss[loss=0.3191, simple_loss=0.358, pruned_loss=0.1401, over 7436.00 frames. ], tot_loss[loss=0.3628, simple_loss=0.4002, pruned_loss=0.1627, over 1616899.10 frames. ], batch size: 17, lr: 2.79e-02, grad_scale: 8.0 +2023-02-05 20:47:16,621 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-05 20:47:23,547 INFO [train.py:901] (1/4) Epoch 2, batch 6600, loss[loss=0.3543, simple_loss=0.3991, pruned_loss=0.1547, over 8507.00 frames. ], tot_loss[loss=0.3629, simple_loss=0.401, pruned_loss=0.1624, over 1619206.31 frames. ], batch size: 26, lr: 2.78e-02, grad_scale: 8.0 +2023-02-05 20:47:36,560 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-05 20:47:45,891 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.289e+02 3.681e+02 4.457e+02 5.556e+02 1.208e+03, threshold=8.913e+02, percent-clipped=4.0 +2023-02-05 20:47:58,950 INFO [train.py:901] (1/4) Epoch 2, batch 6650, loss[loss=0.3458, simple_loss=0.3927, pruned_loss=0.1494, over 7964.00 frames. ], tot_loss[loss=0.3611, simple_loss=0.3997, pruned_loss=0.1613, over 1616754.67 frames. ], batch size: 21, lr: 2.78e-02, grad_scale: 8.0 +2023-02-05 20:48:16,405 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14758.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:48:28,651 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14775.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:48:34,790 INFO [train.py:901] (1/4) Epoch 2, batch 6700, loss[loss=0.3601, simple_loss=0.376, pruned_loss=0.1721, over 7513.00 frames. ], tot_loss[loss=0.3589, simple_loss=0.3979, pruned_loss=0.16, over 1614849.94 frames. ], batch size: 18, lr: 2.78e-02, grad_scale: 8.0 +2023-02-05 20:48:50,186 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14805.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:48:56,674 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 3.873e+02 4.634e+02 6.203e+02 1.536e+03, threshold=9.268e+02, percent-clipped=6.0 +2023-02-05 20:48:58,888 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0539, 3.7586, 2.4966, 2.4945, 3.0461, 2.1885, 2.3598, 2.8989], + device='cuda:1'), covar=tensor([0.1278, 0.0505, 0.0767, 0.0848, 0.0668, 0.0997, 0.1008, 0.0836], + device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0244, 0.0365, 0.0319, 0.0357, 0.0334, 0.0356, 0.0328], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-02-05 20:49:07,140 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14830.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:10,347 INFO [train.py:901] (1/4) Epoch 2, batch 6750, loss[loss=0.3488, simple_loss=0.3991, pruned_loss=0.1492, over 8104.00 frames. ], tot_loss[loss=0.3596, simple_loss=0.3986, pruned_loss=0.1603, over 1613181.27 frames. ], batch size: 23, lr: 2.77e-02, grad_scale: 8.0 +2023-02-05 20:49:11,909 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14836.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:21,254 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14849.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:24,074 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14853.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 20:49:29,653 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14861.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:31,025 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3600, 1.7037, 1.9345, 1.3788, 0.9765, 1.9973, 0.2626, 0.9190], + device='cuda:1'), covar=tensor([0.1821, 0.1049, 0.0768, 0.1376, 0.2379, 0.0604, 0.4357, 0.2108], + device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0100, 0.0084, 0.0142, 0.0144, 0.0082, 0.0163, 0.0116], + device='cuda:1'), out_proj_covar=tensor([1.1963e-04, 1.1657e-04, 9.2294e-05, 1.4965e-04, 1.5701e-04, 9.2883e-05, + 1.7264e-04, 1.3059e-04], device='cuda:1') +2023-02-05 20:49:35,149 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3216, 1.6198, 4.3327, 2.1439, 2.0402, 5.1567, 4.7711, 4.5911], + device='cuda:1'), covar=tensor([0.0972, 0.1288, 0.0169, 0.1406, 0.0755, 0.0151, 0.0202, 0.0312], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0257, 0.0180, 0.0255, 0.0195, 0.0162, 0.0150, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 20:49:41,499 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14878.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:49:45,986 INFO [train.py:901] (1/4) Epoch 2, batch 6800, loss[loss=0.3576, simple_loss=0.4007, pruned_loss=0.1573, over 8107.00 frames. ], tot_loss[loss=0.3595, simple_loss=0.3984, pruned_loss=0.1603, over 1611143.94 frames. ], batch size: 23, lr: 2.77e-02, grad_scale: 8.0 +2023-02-05 20:49:50,353 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14890.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:54,304 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-05 20:50:07,684 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 3.663e+02 4.715e+02 6.092e+02 1.805e+03, threshold=9.431e+02, percent-clipped=7.0 +2023-02-05 20:50:21,320 INFO [train.py:901] (1/4) Epoch 2, batch 6850, loss[loss=0.2994, simple_loss=0.3355, pruned_loss=0.1317, over 7706.00 frames. ], tot_loss[loss=0.3588, simple_loss=0.3984, pruned_loss=0.1596, over 1612556.23 frames. ], batch size: 18, lr: 2.76e-02, grad_scale: 8.0 +2023-02-05 20:50:29,479 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 +2023-02-05 20:50:42,738 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14964.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:50:45,346 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-05 20:50:57,101 INFO [train.py:901] (1/4) Epoch 2, batch 6900, loss[loss=0.3796, simple_loss=0.4084, pruned_loss=0.1754, over 8142.00 frames. ], tot_loss[loss=0.36, simple_loss=0.3995, pruned_loss=0.1602, over 1615129.41 frames. ], batch size: 22, lr: 2.76e-02, grad_scale: 8.0 +2023-02-05 20:51:19,285 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.011e+02 4.191e+02 5.097e+02 7.005e+02 1.700e+03, threshold=1.019e+03, percent-clipped=5.0 +2023-02-05 20:51:32,588 INFO [train.py:901] (1/4) Epoch 2, batch 6950, loss[loss=0.3163, simple_loss=0.3559, pruned_loss=0.1384, over 7249.00 frames. ], tot_loss[loss=0.359, simple_loss=0.3983, pruned_loss=0.1599, over 1610754.68 frames. ], batch size: 16, lr: 2.75e-02, grad_scale: 8.0 +2023-02-05 20:51:56,470 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-05 20:51:57,426 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4890, 1.6819, 2.8439, 1.0166, 2.0850, 1.6384, 1.4470, 1.7130], + device='cuda:1'), covar=tensor([0.1101, 0.1342, 0.0453, 0.2056, 0.1036, 0.1868, 0.1040, 0.1437], + device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0389, 0.0449, 0.0458, 0.0513, 0.0465, 0.0403, 0.0511], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 20:52:08,409 INFO [train.py:901] (1/4) Epoch 2, batch 7000, loss[loss=0.3488, simple_loss=0.3871, pruned_loss=0.1553, over 8085.00 frames. ], tot_loss[loss=0.3586, simple_loss=0.3981, pruned_loss=0.1595, over 1610516.24 frames. ], batch size: 21, lr: 2.75e-02, grad_scale: 8.0 +2023-02-05 20:52:21,504 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15102.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:52:30,568 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.233e+02 3.928e+02 4.810e+02 5.818e+02 1.410e+03, threshold=9.621e+02, percent-clipped=1.0 +2023-02-05 20:52:44,331 INFO [train.py:901] (1/4) Epoch 2, batch 7050, loss[loss=0.3655, simple_loss=0.4229, pruned_loss=0.154, over 8544.00 frames. ], tot_loss[loss=0.3586, simple_loss=0.3983, pruned_loss=0.1594, over 1610928.27 frames. ], batch size: 31, lr: 2.75e-02, grad_scale: 16.0 +2023-02-05 20:52:52,890 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15146.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:53:10,283 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15171.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:53:18,777 INFO [train.py:901] (1/4) Epoch 2, batch 7100, loss[loss=0.3731, simple_loss=0.4078, pruned_loss=0.1692, over 8469.00 frames. ], tot_loss[loss=0.3564, simple_loss=0.3968, pruned_loss=0.158, over 1612410.13 frames. ], batch size: 25, lr: 2.74e-02, grad_scale: 16.0 +2023-02-05 20:53:19,708 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0948, 3.6418, 2.0225, 2.7058, 3.2486, 2.3611, 2.4262, 2.6732], + device='cuda:1'), covar=tensor([0.1487, 0.0641, 0.0978, 0.0881, 0.0541, 0.0992, 0.1261, 0.1023], + device='cuda:1'), in_proj_covar=tensor([0.0371, 0.0246, 0.0349, 0.0314, 0.0347, 0.0335, 0.0353, 0.0322], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-02-05 20:53:39,789 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15213.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:53:40,183 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-05 20:53:41,005 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.137e+02 3.718e+02 4.413e+02 5.855e+02 1.165e+03, threshold=8.826e+02, percent-clipped=3.0 +2023-02-05 20:53:41,246 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8191, 2.0360, 2.4019, 0.9879, 2.6361, 1.7606, 1.2257, 1.9075], + device='cuda:1'), covar=tensor([0.0141, 0.0075, 0.0093, 0.0148, 0.0084, 0.0196, 0.0201, 0.0099], + device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0132, 0.0121, 0.0184, 0.0134, 0.0250, 0.0201, 0.0171], + device='cuda:1'), out_proj_covar=tensor([1.1171e-04, 7.4086e-05, 7.0120e-05, 1.0115e-04, 7.9218e-05, 1.5196e-04, + 1.1549e-04, 9.7090e-05], device='cuda:1') +2023-02-05 20:53:42,519 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15217.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:53:44,510 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:53:53,628 INFO [train.py:901] (1/4) Epoch 2, batch 7150, loss[loss=0.3232, simple_loss=0.3718, pruned_loss=0.1373, over 7926.00 frames. ], tot_loss[loss=0.3557, simple_loss=0.396, pruned_loss=0.1577, over 1611723.67 frames. ], batch size: 20, lr: 2.74e-02, grad_scale: 16.0 +2023-02-05 20:54:02,061 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15245.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:54:29,186 INFO [train.py:901] (1/4) Epoch 2, batch 7200, loss[loss=0.3733, simple_loss=0.3869, pruned_loss=0.1799, over 7259.00 frames. ], tot_loss[loss=0.3556, simple_loss=0.3962, pruned_loss=0.1575, over 1612371.31 frames. ], batch size: 16, lr: 2.73e-02, grad_scale: 16.0 +2023-02-05 20:54:51,169 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.528e+02 3.704e+02 4.905e+02 6.625e+02 1.855e+03, threshold=9.809e+02, percent-clipped=12.0 +2023-02-05 20:55:04,896 INFO [train.py:901] (1/4) Epoch 2, batch 7250, loss[loss=0.3687, simple_loss=0.4213, pruned_loss=0.1581, over 8614.00 frames. ], tot_loss[loss=0.3579, simple_loss=0.3979, pruned_loss=0.1589, over 1611765.69 frames. ], batch size: 31, lr: 2.73e-02, grad_scale: 8.0 +2023-02-05 20:55:21,617 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4353, 2.1744, 2.1561, 0.5318, 2.0971, 1.4501, 0.4259, 1.6750], + device='cuda:1'), covar=tensor([0.0115, 0.0047, 0.0084, 0.0151, 0.0079, 0.0221, 0.0223, 0.0079], + device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0135, 0.0122, 0.0186, 0.0135, 0.0248, 0.0201, 0.0176], + device='cuda:1'), out_proj_covar=tensor([1.1391e-04, 7.5015e-05, 7.0380e-05, 1.0189e-04, 7.9616e-05, 1.5064e-04, + 1.1507e-04, 9.9132e-05], device='cuda:1') +2023-02-05 20:55:22,312 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1638, 1.6002, 1.7789, 1.2430, 0.8860, 1.6456, 0.1127, 0.9047], + device='cuda:1'), covar=tensor([0.2359, 0.1352, 0.0761, 0.1503, 0.2482, 0.0771, 0.4110, 0.2086], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0098, 0.0079, 0.0137, 0.0135, 0.0079, 0.0151, 0.0113], + device='cuda:1'), out_proj_covar=tensor([1.1947e-04, 1.1632e-04, 8.8296e-05, 1.4814e-04, 1.4938e-04, 9.1009e-05, + 1.6417e-04, 1.3109e-04], device='cuda:1') +2023-02-05 20:55:39,924 INFO [train.py:901] (1/4) Epoch 2, batch 7300, loss[loss=0.3527, simple_loss=0.3983, pruned_loss=0.1535, over 8527.00 frames. ], tot_loss[loss=0.3564, simple_loss=0.3961, pruned_loss=0.1583, over 1607423.30 frames. ], batch size: 28, lr: 2.73e-02, grad_scale: 8.0 +2023-02-05 20:56:02,308 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.194e+02 3.434e+02 4.292e+02 5.923e+02 1.449e+03, threshold=8.584e+02, percent-clipped=5.0 +2023-02-05 20:56:12,517 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-05 20:56:14,874 INFO [train.py:901] (1/4) Epoch 2, batch 7350, loss[loss=0.3489, simple_loss=0.3835, pruned_loss=0.1571, over 8232.00 frames. ], tot_loss[loss=0.3581, simple_loss=0.3979, pruned_loss=0.1592, over 1609268.89 frames. ], batch size: 22, lr: 2.72e-02, grad_scale: 8.0 +2023-02-05 20:56:22,685 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-02-05 20:56:24,574 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-05 20:56:42,776 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15473.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:56:43,069 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-05 20:56:43,893 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-05 20:56:49,837 INFO [train.py:901] (1/4) Epoch 2, batch 7400, loss[loss=0.3357, simple_loss=0.3728, pruned_loss=0.1493, over 7268.00 frames. ], tot_loss[loss=0.3588, simple_loss=0.3986, pruned_loss=0.1595, over 1611898.10 frames. ], batch size: 16, lr: 2.72e-02, grad_scale: 8.0 +2023-02-05 20:56:59,509 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15498.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:57:01,972 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-05 20:57:11,806 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 3.940e+02 4.956e+02 6.362e+02 1.377e+03, threshold=9.912e+02, percent-clipped=7.0 +2023-02-05 20:57:24,672 INFO [train.py:901] (1/4) Epoch 2, batch 7450, loss[loss=0.3643, simple_loss=0.4083, pruned_loss=0.1601, over 8352.00 frames. ], tot_loss[loss=0.3593, simple_loss=0.3984, pruned_loss=0.1601, over 1611337.55 frames. ], batch size: 24, lr: 2.71e-02, grad_scale: 8.0 +2023-02-05 20:57:25,215 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.41 vs. limit=5.0 +2023-02-05 20:57:32,760 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9564, 2.0138, 1.6861, 2.6109, 1.1957, 1.1235, 1.3797, 2.1188], + device='cuda:1'), covar=tensor([0.1296, 0.1411, 0.1863, 0.0552, 0.2203, 0.2951, 0.2519, 0.1297], + device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0326, 0.0331, 0.0226, 0.0311, 0.0332, 0.0363, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 20:57:40,474 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15557.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:57:41,767 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-05 20:57:49,569 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.32 vs. limit=5.0 +2023-02-05 20:57:59,035 INFO [train.py:901] (1/4) Epoch 2, batch 7500, loss[loss=0.2878, simple_loss=0.3479, pruned_loss=0.1138, over 8094.00 frames. ], tot_loss[loss=0.3593, simple_loss=0.3985, pruned_loss=0.16, over 1611896.74 frames. ], batch size: 21, lr: 2.71e-02, grad_scale: 8.0 +2023-02-05 20:58:20,424 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-02-05 20:58:21,359 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.209e+02 3.662e+02 4.519e+02 5.678e+02 1.466e+03, threshold=9.038e+02, percent-clipped=6.0 +2023-02-05 20:58:34,049 INFO [train.py:901] (1/4) Epoch 2, batch 7550, loss[loss=0.2709, simple_loss=0.3248, pruned_loss=0.1085, over 7708.00 frames. ], tot_loss[loss=0.3566, simple_loss=0.3966, pruned_loss=0.1583, over 1612086.88 frames. ], batch size: 18, lr: 2.71e-02, grad_scale: 8.0 +2023-02-05 20:59:00,928 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15672.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:59:08,581 INFO [train.py:901] (1/4) Epoch 2, batch 7600, loss[loss=0.295, simple_loss=0.3481, pruned_loss=0.121, over 7426.00 frames. ], tot_loss[loss=0.3565, simple_loss=0.3966, pruned_loss=0.1582, over 1612276.77 frames. ], batch size: 17, lr: 2.70e-02, grad_scale: 8.0 +2023-02-05 20:59:31,058 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.044e+02 3.634e+02 4.473e+02 6.191e+02 1.516e+03, threshold=8.946e+02, percent-clipped=5.0 +2023-02-05 20:59:43,083 INFO [train.py:901] (1/4) Epoch 2, batch 7650, loss[loss=0.3941, simple_loss=0.4261, pruned_loss=0.1811, over 8373.00 frames. ], tot_loss[loss=0.3558, simple_loss=0.3961, pruned_loss=0.1578, over 1608780.12 frames. ], batch size: 24, lr: 2.70e-02, grad_scale: 8.0 +2023-02-05 21:00:19,420 INFO [train.py:901] (1/4) Epoch 2, batch 7700, loss[loss=0.3868, simple_loss=0.421, pruned_loss=0.1763, over 8669.00 frames. ], tot_loss[loss=0.3548, simple_loss=0.3958, pruned_loss=0.1569, over 1613321.21 frames. ], batch size: 34, lr: 2.69e-02, grad_scale: 8.0 +2023-02-05 21:00:41,049 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.759e+02 3.880e+02 4.902e+02 6.175e+02 1.322e+03, threshold=9.805e+02, percent-clipped=4.0 +2023-02-05 21:00:42,379 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-02-05 21:00:51,197 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-05 21:00:53,924 INFO [train.py:901] (1/4) Epoch 2, batch 7750, loss[loss=0.4503, simple_loss=0.4561, pruned_loss=0.2223, over 6676.00 frames. ], tot_loss[loss=0.3535, simple_loss=0.395, pruned_loss=0.156, over 1612106.23 frames. ], batch size: 71, lr: 2.69e-02, grad_scale: 8.0 +2023-02-05 21:01:28,164 INFO [train.py:901] (1/4) Epoch 2, batch 7800, loss[loss=0.2775, simple_loss=0.3334, pruned_loss=0.1108, over 7790.00 frames. ], tot_loss[loss=0.3517, simple_loss=0.393, pruned_loss=0.1552, over 1610323.13 frames. ], batch size: 19, lr: 2.69e-02, grad_scale: 8.0 +2023-02-05 21:01:40,336 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15901.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:01:50,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.227e+02 3.569e+02 4.742e+02 5.990e+02 9.896e+02, threshold=9.484e+02, percent-clipped=1.0 +2023-02-05 21:01:58,284 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.3562, 5.5806, 4.8132, 2.0221, 4.7453, 4.9828, 5.0880, 4.2032], + device='cuda:1'), covar=tensor([0.0996, 0.0545, 0.0872, 0.4851, 0.0538, 0.0519, 0.1214, 0.0883], + device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0224, 0.0265, 0.0358, 0.0241, 0.0201, 0.0255, 0.0185], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 21:01:59,060 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15928.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:02:02,915 INFO [train.py:901] (1/4) Epoch 2, batch 7850, loss[loss=0.3354, simple_loss=0.3732, pruned_loss=0.1488, over 7542.00 frames. ], tot_loss[loss=0.351, simple_loss=0.3928, pruned_loss=0.1546, over 1613108.29 frames. ], batch size: 18, lr: 2.68e-02, grad_scale: 8.0 +2023-02-05 21:02:15,681 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15953.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:02:36,232 INFO [train.py:901] (1/4) Epoch 2, batch 7900, loss[loss=0.3592, simple_loss=0.4064, pruned_loss=0.1559, over 8350.00 frames. ], tot_loss[loss=0.3525, simple_loss=0.3943, pruned_loss=0.1554, over 1618334.90 frames. ], batch size: 24, lr: 2.68e-02, grad_scale: 8.0 +2023-02-05 21:02:58,246 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.267e+02 3.808e+02 4.602e+02 5.936e+02 1.299e+03, threshold=9.205e+02, percent-clipped=9.0 +2023-02-05 21:03:00,391 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16019.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:03:10,218 INFO [train.py:901] (1/4) Epoch 2, batch 7950, loss[loss=0.2751, simple_loss=0.3296, pruned_loss=0.1103, over 7223.00 frames. ], tot_loss[loss=0.3534, simple_loss=0.3951, pruned_loss=0.1559, over 1619964.81 frames. ], batch size: 16, lr: 2.68e-02, grad_scale: 8.0 +2023-02-05 21:03:43,334 INFO [train.py:901] (1/4) Epoch 2, batch 8000, loss[loss=0.3209, simple_loss=0.3686, pruned_loss=0.1365, over 8040.00 frames. ], tot_loss[loss=0.3551, simple_loss=0.3964, pruned_loss=0.1569, over 1619806.53 frames. ], batch size: 22, lr: 2.67e-02, grad_scale: 8.0 +2023-02-05 21:03:56,055 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16103.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:04:04,544 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.336e+02 4.123e+02 4.991e+02 6.647e+02 1.461e+03, threshold=9.983e+02, percent-clipped=10.0 +2023-02-05 21:04:16,506 INFO [train.py:901] (1/4) Epoch 2, batch 8050, loss[loss=0.3703, simple_loss=0.3902, pruned_loss=0.1752, over 7539.00 frames. ], tot_loss[loss=0.3554, simple_loss=0.3956, pruned_loss=0.1576, over 1607700.72 frames. ], batch size: 18, lr: 2.67e-02, grad_scale: 8.0 +2023-02-05 21:04:51,387 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-05 21:04:55,122 INFO [train.py:901] (1/4) Epoch 3, batch 0, loss[loss=0.3948, simple_loss=0.4092, pruned_loss=0.1902, over 7691.00 frames. ], tot_loss[loss=0.3948, simple_loss=0.4092, pruned_loss=0.1902, over 7691.00 frames. ], batch size: 18, lr: 2.53e-02, grad_scale: 8.0 +2023-02-05 21:04:55,123 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 21:05:06,956 INFO [train.py:935] (1/4) Epoch 3, validation: loss=0.2731, simple_loss=0.3579, pruned_loss=0.09417, over 944034.00 frames. +2023-02-05 21:05:06,957 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-05 21:05:07,098 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16167.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:05:23,592 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-05 21:05:24,627 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-02-05 21:05:42,765 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.402e+02 4.065e+02 5.070e+02 6.931e+02 1.670e+03, threshold=1.014e+03, percent-clipped=5.0 +2023-02-05 21:05:42,791 INFO [train.py:901] (1/4) Epoch 3, batch 50, loss[loss=0.4258, simple_loss=0.4456, pruned_loss=0.203, over 7703.00 frames. ], tot_loss[loss=0.36, simple_loss=0.3983, pruned_loss=0.1609, over 359706.15 frames. ], batch size: 72, lr: 2.53e-02, grad_scale: 4.0 +2023-02-05 21:05:58,817 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-05 21:06:02,997 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16245.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:06:18,204 INFO [train.py:901] (1/4) Epoch 3, batch 100, loss[loss=0.4071, simple_loss=0.4318, pruned_loss=0.1912, over 8616.00 frames. ], tot_loss[loss=0.3566, simple_loss=0.3972, pruned_loss=0.158, over 640092.39 frames. ], batch size: 39, lr: 2.53e-02, grad_scale: 4.0 +2023-02-05 21:06:18,921 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-05 21:06:49,925 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-05 21:06:53,421 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.291e+02 3.520e+02 4.471e+02 5.811e+02 1.196e+03, threshold=8.942e+02, percent-clipped=3.0 +2023-02-05 21:06:53,441 INFO [train.py:901] (1/4) Epoch 3, batch 150, loss[loss=0.3913, simple_loss=0.439, pruned_loss=0.1718, over 8242.00 frames. ], tot_loss[loss=0.3535, simple_loss=0.396, pruned_loss=0.1555, over 861618.12 frames. ], batch size: 24, lr: 2.52e-02, grad_scale: 4.0 +2023-02-05 21:07:23,171 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16360.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:07:24,939 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16363.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:07:27,430 INFO [train.py:901] (1/4) Epoch 3, batch 200, loss[loss=0.3523, simple_loss=0.3837, pruned_loss=0.1605, over 7530.00 frames. ], tot_loss[loss=0.3495, simple_loss=0.3934, pruned_loss=0.1528, over 1031012.31 frames. ], batch size: 18, lr: 2.52e-02, grad_scale: 4.0 +2023-02-05 21:07:42,148 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16389.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:07:47,434 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7713, 2.4038, 2.0581, 2.7185, 1.4680, 1.0965, 1.9769, 2.3905], + device='cuda:1'), covar=tensor([0.1264, 0.1180, 0.1427, 0.0502, 0.1826, 0.2713, 0.1677, 0.0955], + device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0329, 0.0329, 0.0233, 0.0306, 0.0329, 0.0354, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 21:08:00,665 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-02-05 21:08:01,470 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 3.609e+02 4.419e+02 5.456e+02 1.161e+03, threshold=8.837e+02, percent-clipped=3.0 +2023-02-05 21:08:01,490 INFO [train.py:901] (1/4) Epoch 3, batch 250, loss[loss=0.3598, simple_loss=0.4128, pruned_loss=0.1534, over 8487.00 frames. ], tot_loss[loss=0.349, simple_loss=0.3929, pruned_loss=0.1525, over 1163351.46 frames. ], batch size: 28, lr: 2.51e-02, grad_scale: 4.0 +2023-02-05 21:08:03,189 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-02-05 21:08:13,923 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-05 21:08:22,557 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-05 21:08:22,621 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16447.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:08:35,579 INFO [train.py:901] (1/4) Epoch 3, batch 300, loss[loss=0.3342, simple_loss=0.3935, pruned_loss=0.1375, over 8466.00 frames. ], tot_loss[loss=0.3487, simple_loss=0.3922, pruned_loss=0.1526, over 1265268.01 frames. ], batch size: 25, lr: 2.51e-02, grad_scale: 4.0 +2023-02-05 21:08:43,564 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16478.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:09:05,163 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16511.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:09:09,099 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 3.752e+02 4.774e+02 5.919e+02 1.248e+03, threshold=9.549e+02, percent-clipped=6.0 +2023-02-05 21:09:09,120 INFO [train.py:901] (1/4) Epoch 3, batch 350, loss[loss=0.3159, simple_loss=0.375, pruned_loss=0.1284, over 8515.00 frames. ], tot_loss[loss=0.349, simple_loss=0.3928, pruned_loss=0.1526, over 1346618.60 frames. ], batch size: 50, lr: 2.51e-02, grad_scale: 4.0 +2023-02-05 21:09:40,929 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16562.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:09:40,987 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5340, 2.1261, 3.5750, 1.0570, 2.6216, 2.0476, 1.6346, 2.1973], + device='cuda:1'), covar=tensor([0.1077, 0.1195, 0.0385, 0.2123, 0.0926, 0.1393, 0.0923, 0.1471], + device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0382, 0.0444, 0.0469, 0.0518, 0.0452, 0.0407, 0.0510], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:09:44,010 INFO [train.py:901] (1/4) Epoch 3, batch 400, loss[loss=0.3245, simple_loss=0.3851, pruned_loss=0.1319, over 8473.00 frames. ], tot_loss[loss=0.3494, simple_loss=0.3932, pruned_loss=0.1528, over 1405991.94 frames. ], batch size: 29, lr: 2.50e-02, grad_scale: 8.0 +2023-02-05 21:10:18,113 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16616.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:10:18,544 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.210e+02 3.588e+02 4.493e+02 6.059e+02 1.047e+03, threshold=8.987e+02, percent-clipped=2.0 +2023-02-05 21:10:18,565 INFO [train.py:901] (1/4) Epoch 3, batch 450, loss[loss=0.3703, simple_loss=0.4142, pruned_loss=0.1632, over 8835.00 frames. ], tot_loss[loss=0.3493, simple_loss=0.3932, pruned_loss=0.1527, over 1458569.05 frames. ], batch size: 32, lr: 2.50e-02, grad_scale: 8.0 +2023-02-05 21:10:21,409 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4596, 2.7391, 3.1844, 2.2084, 1.7280, 3.1346, 0.7001, 1.5902], + device='cuda:1'), covar=tensor([0.2299, 0.2109, 0.0555, 0.1921, 0.3163, 0.0604, 0.5168, 0.1952], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0097, 0.0081, 0.0143, 0.0143, 0.0082, 0.0149, 0.0108], + device='cuda:1'), out_proj_covar=tensor([1.2697e-04, 1.1944e-04, 9.5400e-05, 1.5891e-04, 1.6245e-04, 9.8834e-05, + 1.6792e-04, 1.3025e-04], device='cuda:1') +2023-02-05 21:10:24,806 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16626.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:10:35,583 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16641.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:10:53,050 INFO [train.py:901] (1/4) Epoch 3, batch 500, loss[loss=0.3159, simple_loss=0.3531, pruned_loss=0.1393, over 7541.00 frames. ], tot_loss[loss=0.3497, simple_loss=0.3928, pruned_loss=0.1533, over 1491541.25 frames. ], batch size: 18, lr: 2.50e-02, grad_scale: 8.0 +2023-02-05 21:11:27,933 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 3.547e+02 4.664e+02 6.145e+02 2.246e+03, threshold=9.327e+02, percent-clipped=7.0 +2023-02-05 21:11:27,953 INFO [train.py:901] (1/4) Epoch 3, batch 550, loss[loss=0.364, simple_loss=0.4137, pruned_loss=0.1572, over 8534.00 frames. ], tot_loss[loss=0.3491, simple_loss=0.3926, pruned_loss=0.1528, over 1522533.86 frames. ], batch size: 28, lr: 2.49e-02, grad_scale: 8.0 +2023-02-05 21:11:38,641 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16733.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:11:39,420 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16734.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:11:56,567 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16759.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:12:01,687 INFO [train.py:901] (1/4) Epoch 3, batch 600, loss[loss=0.4002, simple_loss=0.4262, pruned_loss=0.1871, over 8628.00 frames. ], tot_loss[loss=0.3509, simple_loss=0.3932, pruned_loss=0.1543, over 1540715.97 frames. ], batch size: 34, lr: 2.49e-02, grad_scale: 8.0 +2023-02-05 21:12:16,351 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-05 21:12:28,543 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 +2023-02-05 21:12:36,655 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.448e+02 3.715e+02 4.834e+02 5.984e+02 1.404e+03, threshold=9.668e+02, percent-clipped=7.0 +2023-02-05 21:12:36,676 INFO [train.py:901] (1/4) Epoch 3, batch 650, loss[loss=0.3303, simple_loss=0.3702, pruned_loss=0.1452, over 7683.00 frames. ], tot_loss[loss=0.349, simple_loss=0.3915, pruned_loss=0.1533, over 1554958.75 frames. ], batch size: 18, lr: 2.49e-02, grad_scale: 8.0 +2023-02-05 21:12:37,556 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16818.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:12:54,024 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16843.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:12:55,880 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5182, 2.1941, 2.2739, 0.8715, 2.1050, 1.5281, 0.4587, 1.7498], + device='cuda:1'), covar=tensor([0.0100, 0.0040, 0.0039, 0.0123, 0.0065, 0.0194, 0.0182, 0.0065], + device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0141, 0.0121, 0.0187, 0.0134, 0.0251, 0.0206, 0.0174], + device='cuda:1'), out_proj_covar=tensor([1.0708e-04, 7.3793e-05, 6.4750e-05, 9.6656e-05, 7.3934e-05, 1.4436e-04, + 1.1176e-04, 9.3405e-05], device='cuda:1') +2023-02-05 21:12:57,237 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16848.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:13:10,252 INFO [train.py:901] (1/4) Epoch 3, batch 700, loss[loss=0.4567, simple_loss=0.4659, pruned_loss=0.2238, over 8349.00 frames. ], tot_loss[loss=0.3495, simple_loss=0.3924, pruned_loss=0.1533, over 1572411.67 frames. ], batch size: 24, lr: 2.48e-02, grad_scale: 8.0 +2023-02-05 21:13:20,396 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16882.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:13:23,049 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16886.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:13:38,445 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16907.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:13:44,820 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.086e+02 3.932e+02 4.613e+02 6.231e+02 2.383e+03, threshold=9.225e+02, percent-clipped=5.0 +2023-02-05 21:13:44,841 INFO [train.py:901] (1/4) Epoch 3, batch 750, loss[loss=0.4011, simple_loss=0.4213, pruned_loss=0.1905, over 7923.00 frames. ], tot_loss[loss=0.3501, simple_loss=0.3924, pruned_loss=0.1539, over 1575922.58 frames. ], batch size: 20, lr: 2.48e-02, grad_scale: 8.0 +2023-02-05 21:13:49,806 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16924.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:13:52,494 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3329, 4.4526, 3.8642, 1.9426, 3.7895, 3.8246, 4.0459, 3.6613], + device='cuda:1'), covar=tensor([0.0909, 0.0557, 0.0832, 0.4449, 0.0661, 0.0749, 0.1285, 0.0750], + device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0222, 0.0267, 0.0356, 0.0250, 0.0201, 0.0256, 0.0183], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 21:13:59,070 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-05 21:14:07,690 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-05 21:14:19,211 INFO [train.py:901] (1/4) Epoch 3, batch 800, loss[loss=0.3345, simple_loss=0.3748, pruned_loss=0.1471, over 7925.00 frames. ], tot_loss[loss=0.347, simple_loss=0.3901, pruned_loss=0.152, over 1585242.65 frames. ], batch size: 20, lr: 2.48e-02, grad_scale: 8.0 +2023-02-05 21:14:53,628 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.211e+02 3.452e+02 4.368e+02 5.287e+02 1.393e+03, threshold=8.735e+02, percent-clipped=4.0 +2023-02-05 21:14:53,648 INFO [train.py:901] (1/4) Epoch 3, batch 850, loss[loss=0.349, simple_loss=0.4042, pruned_loss=0.1469, over 8361.00 frames. ], tot_loss[loss=0.3501, simple_loss=0.3922, pruned_loss=0.154, over 1592790.31 frames. ], batch size: 24, lr: 2.47e-02, grad_scale: 8.0 +2023-02-05 21:14:55,868 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3934, 2.3961, 1.5108, 1.8967, 1.8152, 1.3420, 1.6826, 1.9169], + device='cuda:1'), covar=tensor([0.1300, 0.0337, 0.0984, 0.0728, 0.0870, 0.1076, 0.0998, 0.0891], + device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0246, 0.0341, 0.0309, 0.0350, 0.0309, 0.0355, 0.0321], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 21:15:28,355 INFO [train.py:901] (1/4) Epoch 3, batch 900, loss[loss=0.2768, simple_loss=0.3454, pruned_loss=0.1041, over 7960.00 frames. ], tot_loss[loss=0.3498, simple_loss=0.3919, pruned_loss=0.1538, over 1594450.86 frames. ], batch size: 21, lr: 2.47e-02, grad_scale: 8.0 +2023-02-05 21:15:53,790 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17104.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:16:02,287 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.375e+02 3.695e+02 4.540e+02 5.760e+02 9.795e+02, threshold=9.080e+02, percent-clipped=3.0 +2023-02-05 21:16:02,308 INFO [train.py:901] (1/4) Epoch 3, batch 950, loss[loss=0.3915, simple_loss=0.4206, pruned_loss=0.1813, over 8581.00 frames. ], tot_loss[loss=0.3483, simple_loss=0.3907, pruned_loss=0.1529, over 1596986.62 frames. ], batch size: 39, lr: 2.47e-02, grad_scale: 8.0 +2023-02-05 21:16:10,479 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17129.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:16:12,703 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-05 21:16:25,713 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-05 21:16:36,855 INFO [train.py:901] (1/4) Epoch 3, batch 1000, loss[loss=0.3593, simple_loss=0.3989, pruned_loss=0.1599, over 8355.00 frames. ], tot_loss[loss=0.3491, simple_loss=0.3906, pruned_loss=0.1538, over 1595706.24 frames. ], batch size: 24, lr: 2.46e-02, grad_scale: 8.0 +2023-02-05 21:16:57,954 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-05 21:17:03,572 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17207.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:17:10,142 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.390e+02 4.093e+02 4.952e+02 6.088e+02 1.030e+03, threshold=9.904e+02, percent-clipped=7.0 +2023-02-05 21:17:10,162 INFO [train.py:901] (1/4) Epoch 3, batch 1050, loss[loss=0.4926, simple_loss=0.4942, pruned_loss=0.2455, over 8594.00 frames. ], tot_loss[loss=0.3476, simple_loss=0.39, pruned_loss=0.1526, over 1603945.02 frames. ], batch size: 39, lr: 2.46e-02, grad_scale: 8.0 +2023-02-05 21:17:10,174 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-05 21:17:19,689 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17230.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:17:45,190 INFO [train.py:901] (1/4) Epoch 3, batch 1100, loss[loss=0.3368, simple_loss=0.383, pruned_loss=0.1453, over 7974.00 frames. ], tot_loss[loss=0.3455, simple_loss=0.3887, pruned_loss=0.1511, over 1606378.32 frames. ], batch size: 21, lr: 2.46e-02, grad_scale: 8.0 +2023-02-05 21:17:45,926 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17268.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:17:48,751 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.31 vs. limit=5.0 +2023-02-05 21:18:19,092 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.272e+02 3.840e+02 4.434e+02 5.714e+02 1.415e+03, threshold=8.869e+02, percent-clipped=3.0 +2023-02-05 21:18:19,112 INFO [train.py:901] (1/4) Epoch 3, batch 1150, loss[loss=0.3184, simple_loss=0.3656, pruned_loss=0.1356, over 8247.00 frames. ], tot_loss[loss=0.343, simple_loss=0.3868, pruned_loss=0.1496, over 1602747.90 frames. ], batch size: 22, lr: 2.45e-02, grad_scale: 8.0 +2023-02-05 21:18:22,465 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-05 21:18:22,641 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7217, 2.1653, 1.9335, 2.6711, 1.1672, 1.3046, 1.5295, 2.3158], + device='cuda:1'), covar=tensor([0.1361, 0.1364, 0.1557, 0.0561, 0.2075, 0.2568, 0.2003, 0.1007], + device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0330, 0.0324, 0.0232, 0.0302, 0.0332, 0.0355, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 21:18:35,405 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4930, 1.9339, 3.2058, 1.0234, 2.2727, 1.7185, 1.4810, 1.8950], + device='cuda:1'), covar=tensor([0.1202, 0.1349, 0.0421, 0.2396, 0.1128, 0.1765, 0.1036, 0.1705], + device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0399, 0.0458, 0.0490, 0.0535, 0.0476, 0.0423, 0.0537], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:18:38,632 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17345.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:18:52,840 INFO [train.py:901] (1/4) Epoch 3, batch 1200, loss[loss=0.3181, simple_loss=0.3835, pruned_loss=0.1263, over 8293.00 frames. ], tot_loss[loss=0.342, simple_loss=0.3862, pruned_loss=0.1488, over 1604562.38 frames. ], batch size: 23, lr: 2.45e-02, grad_scale: 8.0 +2023-02-05 21:19:02,187 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17380.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:19:04,907 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17383.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:19:17,665 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17401.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:19:28,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.147e+02 3.772e+02 4.989e+02 5.905e+02 9.785e+02, threshold=9.978e+02, percent-clipped=4.0 +2023-02-05 21:19:28,385 INFO [train.py:901] (1/4) Epoch 3, batch 1250, loss[loss=0.3769, simple_loss=0.4142, pruned_loss=0.1698, over 6847.00 frames. ], tot_loss[loss=0.3419, simple_loss=0.3858, pruned_loss=0.149, over 1602754.96 frames. ], batch size: 71, lr: 2.45e-02, grad_scale: 8.0 +2023-02-05 21:20:02,610 INFO [train.py:901] (1/4) Epoch 3, batch 1300, loss[loss=0.3043, simple_loss=0.3577, pruned_loss=0.1254, over 8196.00 frames. ], tot_loss[loss=0.3438, simple_loss=0.3879, pruned_loss=0.1498, over 1609980.42 frames. ], batch size: 23, lr: 2.44e-02, grad_scale: 8.0 +2023-02-05 21:20:37,546 INFO [train.py:901] (1/4) Epoch 3, batch 1350, loss[loss=0.3288, simple_loss=0.3958, pruned_loss=0.1309, over 8614.00 frames. ], tot_loss[loss=0.3455, simple_loss=0.389, pruned_loss=0.151, over 1613145.39 frames. ], batch size: 34, lr: 2.44e-02, grad_scale: 4.0 +2023-02-05 21:20:38,234 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.198e+02 4.258e+02 5.812e+02 8.345e+02 8.746e+03, threshold=1.162e+03, percent-clipped=16.0 +2023-02-05 21:21:00,247 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17551.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:21:11,045 INFO [train.py:901] (1/4) Epoch 3, batch 1400, loss[loss=0.3668, simple_loss=0.4119, pruned_loss=0.1609, over 8511.00 frames. ], tot_loss[loss=0.3463, simple_loss=0.3899, pruned_loss=0.1513, over 1616018.21 frames. ], batch size: 28, lr: 2.44e-02, grad_scale: 4.0 +2023-02-05 21:21:34,754 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17601.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:21:38,919 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-02-05 21:21:46,907 INFO [train.py:901] (1/4) Epoch 3, batch 1450, loss[loss=0.3337, simple_loss=0.384, pruned_loss=0.1417, over 8148.00 frames. ], tot_loss[loss=0.347, simple_loss=0.3906, pruned_loss=0.1517, over 1618377.58 frames. ], batch size: 22, lr: 2.43e-02, grad_scale: 4.0 +2023-02-05 21:21:47,583 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 3.309e+02 4.161e+02 5.035e+02 1.114e+03, threshold=8.322e+02, percent-clipped=0.0 +2023-02-05 21:21:48,940 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-05 21:21:53,237 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17626.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:22:02,518 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17639.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:22:19,049 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17664.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:22:20,439 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17666.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:22:20,943 INFO [train.py:901] (1/4) Epoch 3, batch 1500, loss[loss=0.3153, simple_loss=0.372, pruned_loss=0.1293, over 8316.00 frames. ], tot_loss[loss=0.3476, simple_loss=0.391, pruned_loss=0.1521, over 1617228.19 frames. ], batch size: 25, lr: 2.43e-02, grad_scale: 4.0 +2023-02-05 21:22:56,188 INFO [train.py:901] (1/4) Epoch 3, batch 1550, loss[loss=0.3103, simple_loss=0.3532, pruned_loss=0.1337, over 7800.00 frames. ], tot_loss[loss=0.3489, simple_loss=0.3917, pruned_loss=0.153, over 1615940.74 frames. ], batch size: 19, lr: 2.43e-02, grad_scale: 4.0 +2023-02-05 21:22:56,832 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.415e+02 3.678e+02 4.620e+02 5.892e+02 1.697e+03, threshold=9.239e+02, percent-clipped=9.0 +2023-02-05 21:22:57,018 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17718.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:23:01,087 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17724.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:23:16,062 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17745.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:23:17,955 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17748.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:23:30,930 INFO [train.py:901] (1/4) Epoch 3, batch 1600, loss[loss=0.3095, simple_loss=0.3574, pruned_loss=0.1309, over 7438.00 frames. ], tot_loss[loss=0.3486, simple_loss=0.3916, pruned_loss=0.1528, over 1616090.29 frames. ], batch size: 17, lr: 2.42e-02, grad_scale: 8.0 +2023-02-05 21:23:46,407 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2244, 4.3308, 3.7830, 1.6393, 3.7450, 3.6833, 4.0509, 3.3416], + device='cuda:1'), covar=tensor([0.0812, 0.0400, 0.0722, 0.4390, 0.0512, 0.0679, 0.0764, 0.0656], + device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0231, 0.0278, 0.0364, 0.0253, 0.0207, 0.0263, 0.0194], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 21:23:49,220 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6146, 2.2526, 2.3261, 0.4633, 2.2994, 1.3838, 0.5981, 1.9221], + device='cuda:1'), covar=tensor([0.0137, 0.0054, 0.0084, 0.0190, 0.0091, 0.0262, 0.0222, 0.0073], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0151, 0.0125, 0.0189, 0.0143, 0.0262, 0.0209, 0.0178], + device='cuda:1'), out_proj_covar=tensor([1.1028e-04, 7.7879e-05, 6.4757e-05, 9.4771e-05, 7.5891e-05, 1.4564e-04, + 1.0997e-04, 9.3637e-05], device='cuda:1') +2023-02-05 21:24:05,136 INFO [train.py:901] (1/4) Epoch 3, batch 1650, loss[loss=0.3248, simple_loss=0.3859, pruned_loss=0.1318, over 8327.00 frames. ], tot_loss[loss=0.3464, simple_loss=0.39, pruned_loss=0.1514, over 1614778.73 frames. ], batch size: 26, lr: 2.42e-02, grad_scale: 8.0 +2023-02-05 21:24:05,797 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.224e+02 3.309e+02 4.132e+02 5.477e+02 8.650e+02, threshold=8.264e+02, percent-clipped=0.0 +2023-02-05 21:24:20,682 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17839.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:24:26,724 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1338, 2.3634, 1.9964, 2.8588, 1.4166, 1.4482, 1.9802, 2.6334], + device='cuda:1'), covar=tensor([0.1038, 0.1245, 0.1456, 0.0515, 0.1907, 0.2194, 0.1940, 0.1004], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0318, 0.0313, 0.0228, 0.0292, 0.0320, 0.0335, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 21:24:35,267 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17860.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:24:39,691 INFO [train.py:901] (1/4) Epoch 3, batch 1700, loss[loss=0.3146, simple_loss=0.3688, pruned_loss=0.1302, over 8136.00 frames. ], tot_loss[loss=0.3476, simple_loss=0.3913, pruned_loss=0.152, over 1615726.67 frames. ], batch size: 22, lr: 2.42e-02, grad_scale: 8.0 +2023-02-05 21:25:13,899 INFO [train.py:901] (1/4) Epoch 3, batch 1750, loss[loss=0.3466, simple_loss=0.3859, pruned_loss=0.1536, over 7783.00 frames. ], tot_loss[loss=0.3493, simple_loss=0.393, pruned_loss=0.1528, over 1620838.47 frames. ], batch size: 19, lr: 2.42e-02, grad_scale: 8.0 +2023-02-05 21:25:14,599 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.106e+02 3.998e+02 5.161e+02 6.686e+02 1.470e+03, threshold=1.032e+03, percent-clipped=12.0 +2023-02-05 21:25:17,619 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17922.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:25:35,790 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17947.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:25:48,900 INFO [train.py:901] (1/4) Epoch 3, batch 1800, loss[loss=0.2733, simple_loss=0.3198, pruned_loss=0.1134, over 7701.00 frames. ], tot_loss[loss=0.3462, simple_loss=0.3904, pruned_loss=0.151, over 1618815.48 frames. ], batch size: 18, lr: 2.41e-02, grad_scale: 8.0 +2023-02-05 21:25:57,026 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17978.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:26:25,060 INFO [train.py:901] (1/4) Epoch 3, batch 1850, loss[loss=0.3552, simple_loss=0.4127, pruned_loss=0.1489, over 8478.00 frames. ], tot_loss[loss=0.3454, simple_loss=0.3897, pruned_loss=0.1506, over 1619211.08 frames. ], batch size: 28, lr: 2.41e-02, grad_scale: 8.0 +2023-02-05 21:26:25,631 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.340e+02 3.564e+02 4.327e+02 5.819e+02 2.228e+03, threshold=8.654e+02, percent-clipped=8.0 +2023-02-05 21:26:39,909 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7495, 2.1597, 1.8650, 2.6629, 1.2440, 1.3037, 1.6110, 2.3149], + device='cuda:1'), covar=tensor([0.1333, 0.1535, 0.1669, 0.0485, 0.1944, 0.2436, 0.2080, 0.0922], + device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0316, 0.0320, 0.0234, 0.0295, 0.0327, 0.0348, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 21:26:52,697 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1963, 1.8047, 2.9979, 2.5107, 2.3098, 1.9945, 1.3710, 1.1573], + device='cuda:1'), covar=tensor([0.0721, 0.0805, 0.0166, 0.0289, 0.0319, 0.0375, 0.0531, 0.0768], + device='cuda:1'), in_proj_covar=tensor([0.0561, 0.0487, 0.0387, 0.0435, 0.0536, 0.0452, 0.0474, 0.0478], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 21:26:55,935 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18062.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:26:59,914 INFO [train.py:901] (1/4) Epoch 3, batch 1900, loss[loss=0.4159, simple_loss=0.4414, pruned_loss=0.1953, over 8511.00 frames. ], tot_loss[loss=0.3462, simple_loss=0.3904, pruned_loss=0.1509, over 1621279.56 frames. ], batch size: 49, lr: 2.41e-02, grad_scale: 8.0 +2023-02-05 21:27:15,649 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.17 vs. limit=5.0 +2023-02-05 21:27:17,421 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18092.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:27:19,615 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18095.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:27:24,179 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-05 21:27:34,371 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18116.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:27:34,803 INFO [train.py:901] (1/4) Epoch 3, batch 1950, loss[loss=0.4042, simple_loss=0.444, pruned_loss=0.1822, over 8614.00 frames. ], tot_loss[loss=0.3445, simple_loss=0.3894, pruned_loss=0.1499, over 1617004.96 frames. ], batch size: 34, lr: 2.40e-02, grad_scale: 8.0 +2023-02-05 21:27:35,483 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.131e+02 3.385e+02 4.094e+02 5.586e+02 1.173e+03, threshold=8.188e+02, percent-clipped=3.0 +2023-02-05 21:27:36,216 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-05 21:27:37,024 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18120.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:27:51,206 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18141.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:27:55,040 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-05 21:28:09,123 INFO [train.py:901] (1/4) Epoch 3, batch 2000, loss[loss=0.4401, simple_loss=0.4632, pruned_loss=0.2085, over 8332.00 frames. ], tot_loss[loss=0.3422, simple_loss=0.3875, pruned_loss=0.1484, over 1617970.58 frames. ], batch size: 25, lr: 2.40e-02, grad_scale: 8.0 +2023-02-05 21:28:17,060 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18177.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:28:27,346 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18192.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:28:28,019 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18193.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:28:38,158 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18207.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:28:41,236 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-02-05 21:28:44,807 INFO [train.py:901] (1/4) Epoch 3, batch 2050, loss[loss=0.2794, simple_loss=0.3349, pruned_loss=0.112, over 7937.00 frames. ], tot_loss[loss=0.3418, simple_loss=0.3873, pruned_loss=0.1482, over 1620512.70 frames. ], batch size: 20, lr: 2.40e-02, grad_scale: 8.0 +2023-02-05 21:28:46,138 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.474e+02 3.817e+02 4.995e+02 6.129e+02 1.664e+03, threshold=9.991e+02, percent-clipped=7.0 +2023-02-05 21:29:03,574 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4595, 2.6969, 1.6806, 2.0817, 2.0384, 1.2197, 1.7656, 2.1845], + device='cuda:1'), covar=tensor([0.1265, 0.0356, 0.1024, 0.0766, 0.0798, 0.1337, 0.1089, 0.0772], + device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0259, 0.0354, 0.0319, 0.0352, 0.0317, 0.0363, 0.0322], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 21:29:19,897 INFO [train.py:901] (1/4) Epoch 3, batch 2100, loss[loss=0.2879, simple_loss=0.3474, pruned_loss=0.1142, over 8094.00 frames. ], tot_loss[loss=0.3433, simple_loss=0.3884, pruned_loss=0.1491, over 1622054.82 frames. ], batch size: 21, lr: 2.39e-02, grad_scale: 8.0 +2023-02-05 21:29:55,202 INFO [train.py:901] (1/4) Epoch 3, batch 2150, loss[loss=0.3449, simple_loss=0.4012, pruned_loss=0.1443, over 8445.00 frames. ], tot_loss[loss=0.3422, simple_loss=0.3873, pruned_loss=0.1486, over 1620460.51 frames. ], batch size: 27, lr: 2.39e-02, grad_scale: 8.0 +2023-02-05 21:29:55,885 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.297e+02 3.744e+02 4.718e+02 5.936e+02 1.452e+03, threshold=9.436e+02, percent-clipped=4.0 +2023-02-05 21:29:59,210 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18322.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:30:15,995 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1247, 1.1508, 2.3221, 1.0191, 2.0395, 2.5245, 2.3480, 2.1272], + device='cuda:1'), covar=tensor([0.1034, 0.1118, 0.0397, 0.1852, 0.0480, 0.0264, 0.0328, 0.0587], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0252, 0.0196, 0.0255, 0.0196, 0.0162, 0.0162, 0.0238], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 21:30:31,074 INFO [train.py:901] (1/4) Epoch 3, batch 2200, loss[loss=0.4405, simple_loss=0.4435, pruned_loss=0.2188, over 6923.00 frames. ], tot_loss[loss=0.3423, simple_loss=0.387, pruned_loss=0.1488, over 1615380.93 frames. ], batch size: 71, lr: 2.39e-02, grad_scale: 8.0 +2023-02-05 21:31:06,840 INFO [train.py:901] (1/4) Epoch 3, batch 2250, loss[loss=0.3263, simple_loss=0.386, pruned_loss=0.1333, over 8508.00 frames. ], tot_loss[loss=0.341, simple_loss=0.3864, pruned_loss=0.1478, over 1614600.05 frames. ], batch size: 26, lr: 2.38e-02, grad_scale: 8.0 +2023-02-05 21:31:07,502 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.379e+02 3.424e+02 4.222e+02 5.561e+02 1.530e+03, threshold=8.445e+02, percent-clipped=2.0 +2023-02-05 21:31:18,180 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18433.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:31:21,447 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18437.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:31:36,360 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18458.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:31:39,724 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7157, 1.4654, 3.0438, 1.4117, 2.3470, 3.4285, 3.0857, 2.9639], + device='cuda:1'), covar=tensor([0.1311, 0.1557, 0.0501, 0.1924, 0.0701, 0.0282, 0.0356, 0.0557], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0260, 0.0199, 0.0255, 0.0200, 0.0165, 0.0168, 0.0245], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 21:31:39,795 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18463.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:31:42,290 INFO [train.py:901] (1/4) Epoch 3, batch 2300, loss[loss=0.2855, simple_loss=0.3458, pruned_loss=0.1126, over 7523.00 frames. ], tot_loss[loss=0.3403, simple_loss=0.3861, pruned_loss=0.1472, over 1615795.61 frames. ], batch size: 18, lr: 2.38e-02, grad_scale: 8.0 +2023-02-05 21:31:56,790 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18488.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:32:10,468 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18508.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:32:15,216 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7380, 1.5301, 3.3501, 1.1812, 2.2420, 3.8996, 3.6146, 3.2845], + device='cuda:1'), covar=tensor([0.1222, 0.1468, 0.0328, 0.2043, 0.0638, 0.0227, 0.0289, 0.0520], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0261, 0.0199, 0.0255, 0.0201, 0.0166, 0.0168, 0.0245], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 21:32:17,121 INFO [train.py:901] (1/4) Epoch 3, batch 2350, loss[loss=0.3182, simple_loss=0.3639, pruned_loss=0.1362, over 7449.00 frames. ], tot_loss[loss=0.3383, simple_loss=0.385, pruned_loss=0.1458, over 1617256.44 frames. ], batch size: 17, lr: 2.38e-02, grad_scale: 8.0 +2023-02-05 21:32:17,770 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.449e+02 3.759e+02 4.661e+02 5.652e+02 9.227e+02, threshold=9.323e+02, percent-clipped=1.0 +2023-02-05 21:32:23,347 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18526.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:32:30,484 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18536.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:32:31,167 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18537.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:32:51,324 INFO [train.py:901] (1/4) Epoch 3, batch 2400, loss[loss=0.3082, simple_loss=0.3544, pruned_loss=0.131, over 7921.00 frames. ], tot_loss[loss=0.339, simple_loss=0.3853, pruned_loss=0.1464, over 1614788.01 frames. ], batch size: 20, lr: 2.38e-02, grad_scale: 8.0 +2023-02-05 21:33:05,630 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2023, 1.1875, 4.3725, 1.7979, 3.6526, 3.6107, 3.7647, 3.7302], + device='cuda:1'), covar=tensor([0.0279, 0.3267, 0.0220, 0.1787, 0.0907, 0.0411, 0.0394, 0.0456], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0414, 0.0279, 0.0320, 0.0383, 0.0302, 0.0290, 0.0325], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 21:33:24,822 INFO [train.py:901] (1/4) Epoch 3, batch 2450, loss[loss=0.3928, simple_loss=0.4202, pruned_loss=0.1827, over 8198.00 frames. ], tot_loss[loss=0.343, simple_loss=0.3882, pruned_loss=0.1489, over 1616517.92 frames. ], batch size: 23, lr: 2.37e-02, grad_scale: 8.0 +2023-02-05 21:33:25,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 3.618e+02 4.763e+02 6.456e+02 1.024e+03, threshold=9.527e+02, percent-clipped=2.0 +2023-02-05 21:33:49,164 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18651.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:33:49,837 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18652.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:33:51,269 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7104, 2.4877, 3.7304, 3.3086, 2.7576, 2.2990, 1.4566, 2.0499], + device='cuda:1'), covar=tensor([0.0859, 0.1037, 0.0248, 0.0399, 0.0549, 0.0491, 0.0703, 0.0962], + device='cuda:1'), in_proj_covar=tensor([0.0566, 0.0491, 0.0393, 0.0449, 0.0544, 0.0464, 0.0485, 0.0487], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 21:33:59,429 INFO [train.py:901] (1/4) Epoch 3, batch 2500, loss[loss=0.3672, simple_loss=0.4102, pruned_loss=0.1621, over 8468.00 frames. ], tot_loss[loss=0.3416, simple_loss=0.3874, pruned_loss=0.1478, over 1617920.19 frames. ], batch size: 25, lr: 2.37e-02, grad_scale: 8.0 +2023-02-05 21:34:17,683 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18692.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:34:18,343 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18693.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:34:33,854 INFO [train.py:901] (1/4) Epoch 3, batch 2550, loss[loss=0.3868, simple_loss=0.42, pruned_loss=0.1768, over 8448.00 frames. ], tot_loss[loss=0.3427, simple_loss=0.3884, pruned_loss=0.1484, over 1619287.37 frames. ], batch size: 49, lr: 2.37e-02, grad_scale: 8.0 +2023-02-05 21:34:34,503 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.095e+02 3.889e+02 4.529e+02 5.619e+02 1.309e+03, threshold=9.058e+02, percent-clipped=5.0 +2023-02-05 21:34:34,734 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18718.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:34:40,293 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-02-05 21:35:00,848 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-05 21:35:08,024 INFO [train.py:901] (1/4) Epoch 3, batch 2600, loss[loss=0.3472, simple_loss=0.3878, pruned_loss=0.1533, over 8647.00 frames. ], tot_loss[loss=0.3393, simple_loss=0.3858, pruned_loss=0.1464, over 1618778.95 frames. ], batch size: 34, lr: 2.36e-02, grad_scale: 8.0 +2023-02-05 21:35:27,952 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5890, 1.9767, 3.5715, 0.9915, 2.4403, 1.7694, 1.6226, 2.1030], + device='cuda:1'), covar=tensor([0.1125, 0.1284, 0.0467, 0.2245, 0.1053, 0.1830, 0.1050, 0.1804], + device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0408, 0.0481, 0.0493, 0.0549, 0.0491, 0.0441, 0.0547], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:35:44,452 INFO [train.py:901] (1/4) Epoch 3, batch 2650, loss[loss=0.2848, simple_loss=0.3468, pruned_loss=0.1114, over 8251.00 frames. ], tot_loss[loss=0.3387, simple_loss=0.3855, pruned_loss=0.1459, over 1622709.17 frames. ], batch size: 22, lr: 2.36e-02, grad_scale: 8.0 +2023-02-05 21:35:45,137 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.448e+02 3.426e+02 4.272e+02 5.708e+02 1.020e+03, threshold=8.544e+02, percent-clipped=5.0 +2023-02-05 21:36:08,385 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18852.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:36:19,356 INFO [train.py:901] (1/4) Epoch 3, batch 2700, loss[loss=0.4109, simple_loss=0.435, pruned_loss=0.1934, over 7253.00 frames. ], tot_loss[loss=0.339, simple_loss=0.3859, pruned_loss=0.1461, over 1626135.53 frames. ], batch size: 71, lr: 2.36e-02, grad_scale: 8.0 +2023-02-05 21:36:21,520 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18870.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:36:28,573 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-05 21:36:47,163 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18907.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:36:47,843 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18908.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:36:53,561 INFO [train.py:901] (1/4) Epoch 3, batch 2750, loss[loss=0.3406, simple_loss=0.393, pruned_loss=0.1441, over 8246.00 frames. ], tot_loss[loss=0.3399, simple_loss=0.3865, pruned_loss=0.1466, over 1619618.54 frames. ], batch size: 24, lr: 2.36e-02, grad_scale: 8.0 +2023-02-05 21:36:54,219 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 3.360e+02 4.052e+02 5.079e+02 9.265e+02, threshold=8.105e+02, percent-clipped=2.0 +2023-02-05 21:37:05,002 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18932.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:37:05,663 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18933.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:37:15,752 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18948.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:37:28,079 INFO [train.py:901] (1/4) Epoch 3, batch 2800, loss[loss=0.2667, simple_loss=0.3197, pruned_loss=0.1068, over 7429.00 frames. ], tot_loss[loss=0.3393, simple_loss=0.3864, pruned_loss=0.1461, over 1619959.02 frames. ], batch size: 17, lr: 2.35e-02, grad_scale: 8.0 +2023-02-05 21:37:28,254 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18967.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:37:41,129 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18985.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:38:03,344 INFO [train.py:901] (1/4) Epoch 3, batch 2850, loss[loss=0.33, simple_loss=0.3852, pruned_loss=0.1374, over 7802.00 frames. ], tot_loss[loss=0.3386, simple_loss=0.3861, pruned_loss=0.1455, over 1622326.17 frames. ], batch size: 20, lr: 2.35e-02, grad_scale: 8.0 +2023-02-05 21:38:03,913 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.788e+02 3.511e+02 4.402e+02 5.555e+02 1.104e+03, threshold=8.804e+02, percent-clipped=5.0 +2023-02-05 21:38:15,997 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19036.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:38:37,626 INFO [train.py:901] (1/4) Epoch 3, batch 2900, loss[loss=0.334, simple_loss=0.375, pruned_loss=0.1465, over 7421.00 frames. ], tot_loss[loss=0.3392, simple_loss=0.3863, pruned_loss=0.1461, over 1620707.90 frames. ], batch size: 17, lr: 2.35e-02, grad_scale: 8.0 +2023-02-05 21:38:48,617 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-02-05 21:39:02,486 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-05 21:39:09,997 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2209, 1.7181, 1.5843, 0.3720, 1.5434, 1.1990, 0.3260, 1.6458], + device='cuda:1'), covar=tensor([0.0102, 0.0047, 0.0058, 0.0125, 0.0069, 0.0212, 0.0165, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0148, 0.0135, 0.0197, 0.0151, 0.0272, 0.0212, 0.0182], + device='cuda:1'), out_proj_covar=tensor([1.0782e-04, 7.3127e-05, 6.6019e-05, 9.5798e-05, 7.7379e-05, 1.4546e-04, + 1.0817e-04, 9.1182e-05], device='cuda:1') +2023-02-05 21:39:11,764 INFO [train.py:901] (1/4) Epoch 3, batch 2950, loss[loss=0.3298, simple_loss=0.3773, pruned_loss=0.1412, over 8132.00 frames. ], tot_loss[loss=0.3381, simple_loss=0.3852, pruned_loss=0.1455, over 1617930.99 frames. ], batch size: 22, lr: 2.34e-02, grad_scale: 8.0 +2023-02-05 21:39:12,417 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 3.613e+02 4.498e+02 5.900e+02 1.326e+03, threshold=8.996e+02, percent-clipped=8.0 +2023-02-05 21:39:32,502 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19147.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:39:35,223 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19151.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:39:46,215 INFO [train.py:901] (1/4) Epoch 3, batch 3000, loss[loss=0.3379, simple_loss=0.3845, pruned_loss=0.1456, over 8089.00 frames. ], tot_loss[loss=0.3355, simple_loss=0.383, pruned_loss=0.144, over 1617722.86 frames. ], batch size: 21, lr: 2.34e-02, grad_scale: 8.0 +2023-02-05 21:39:46,215 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 21:39:58,666 INFO [train.py:935] (1/4) Epoch 3, validation: loss=0.2584, simple_loss=0.3473, pruned_loss=0.08481, over 944034.00 frames. +2023-02-05 21:39:58,667 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-05 21:40:03,995 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.85 vs. limit=5.0 +2023-02-05 21:40:29,846 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2464, 3.8425, 2.4066, 2.8391, 2.4866, 1.9002, 2.4488, 2.7378], + device='cuda:1'), covar=tensor([0.1224, 0.0440, 0.0780, 0.0618, 0.0771, 0.1054, 0.0960, 0.1026], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0245, 0.0326, 0.0303, 0.0333, 0.0310, 0.0341, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-02-05 21:40:33,767 INFO [train.py:901] (1/4) Epoch 3, batch 3050, loss[loss=0.3854, simple_loss=0.4106, pruned_loss=0.1801, over 7283.00 frames. ], tot_loss[loss=0.337, simple_loss=0.3841, pruned_loss=0.1449, over 1616948.28 frames. ], batch size: 72, lr: 2.34e-02, grad_scale: 8.0 +2023-02-05 21:40:34,446 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.016e+02 3.526e+02 4.458e+02 6.217e+02 1.354e+03, threshold=8.917e+02, percent-clipped=3.0 +2023-02-05 21:40:38,079 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19223.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:40:38,977 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.81 vs. limit=5.0 +2023-02-05 21:40:50,792 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19241.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:40:55,431 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19248.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:41:07,576 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19266.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:41:07,997 INFO [train.py:901] (1/4) Epoch 3, batch 3100, loss[loss=0.3348, simple_loss=0.3775, pruned_loss=0.146, over 7527.00 frames. ], tot_loss[loss=0.337, simple_loss=0.384, pruned_loss=0.145, over 1616077.19 frames. ], batch size: 18, lr: 2.34e-02, grad_scale: 8.0 +2023-02-05 21:41:26,056 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19292.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:41:43,825 INFO [train.py:901] (1/4) Epoch 3, batch 3150, loss[loss=0.3611, simple_loss=0.3924, pruned_loss=0.1649, over 8248.00 frames. ], tot_loss[loss=0.3383, simple_loss=0.3846, pruned_loss=0.146, over 1615517.80 frames. ], batch size: 22, lr: 2.33e-02, grad_scale: 8.0 +2023-02-05 21:41:44,465 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.556e+02 3.507e+02 4.387e+02 6.193e+02 1.521e+03, threshold=8.773e+02, percent-clipped=4.0 +2023-02-05 21:42:02,338 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.52 vs. limit=5.0 +2023-02-05 21:42:15,246 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0945, 0.9872, 3.1225, 0.9205, 2.6119, 2.6909, 2.8303, 2.7701], + device='cuda:1'), covar=tensor([0.0482, 0.3402, 0.0546, 0.2152, 0.1402, 0.0635, 0.0559, 0.0695], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0415, 0.0287, 0.0325, 0.0387, 0.0308, 0.0302, 0.0326], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 21:42:17,833 INFO [train.py:901] (1/4) Epoch 3, batch 3200, loss[loss=0.3353, simple_loss=0.3932, pruned_loss=0.1387, over 8015.00 frames. ], tot_loss[loss=0.3376, simple_loss=0.384, pruned_loss=0.1456, over 1613430.71 frames. ], batch size: 22, lr: 2.33e-02, grad_scale: 8.0 +2023-02-05 21:42:45,873 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19407.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:42:45,912 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19407.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:42:49,211 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19412.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:42:53,530 INFO [train.py:901] (1/4) Epoch 3, batch 3250, loss[loss=0.4182, simple_loss=0.4449, pruned_loss=0.1958, over 8683.00 frames. ], tot_loss[loss=0.3394, simple_loss=0.3851, pruned_loss=0.1469, over 1610535.75 frames. ], batch size: 30, lr: 2.33e-02, grad_scale: 8.0 +2023-02-05 21:42:54,124 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 3.440e+02 4.583e+02 5.736e+02 1.373e+03, threshold=9.167e+02, percent-clipped=8.0 +2023-02-05 21:42:54,452 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-05 21:43:03,724 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19432.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:43:26,830 INFO [train.py:901] (1/4) Epoch 3, batch 3300, loss[loss=0.2875, simple_loss=0.3313, pruned_loss=0.1219, over 7226.00 frames. ], tot_loss[loss=0.339, simple_loss=0.3847, pruned_loss=0.1466, over 1609076.85 frames. ], batch size: 16, lr: 2.32e-02, grad_scale: 8.0 +2023-02-05 21:43:43,382 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19491.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:44:01,035 INFO [train.py:901] (1/4) Epoch 3, batch 3350, loss[loss=0.3831, simple_loss=0.415, pruned_loss=0.1756, over 8456.00 frames. ], tot_loss[loss=0.3408, simple_loss=0.3859, pruned_loss=0.1479, over 1612450.08 frames. ], batch size: 48, lr: 2.32e-02, grad_scale: 16.0 +2023-02-05 21:44:01,702 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 3.690e+02 4.650e+02 5.581e+02 1.223e+03, threshold=9.300e+02, percent-clipped=5.0 +2023-02-05 21:44:35,822 INFO [train.py:901] (1/4) Epoch 3, batch 3400, loss[loss=0.3757, simple_loss=0.4178, pruned_loss=0.1668, over 8550.00 frames. ], tot_loss[loss=0.3401, simple_loss=0.3855, pruned_loss=0.1474, over 1612101.21 frames. ], batch size: 39, lr: 2.32e-02, grad_scale: 16.0 +2023-02-05 21:45:02,606 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19606.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:45:09,812 INFO [train.py:901] (1/4) Epoch 3, batch 3450, loss[loss=0.375, simple_loss=0.42, pruned_loss=0.1651, over 8763.00 frames. ], tot_loss[loss=0.3397, simple_loss=0.385, pruned_loss=0.1472, over 1614701.79 frames. ], batch size: 30, lr: 2.32e-02, grad_scale: 16.0 +2023-02-05 21:45:10,436 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.543e+02 3.801e+02 4.733e+02 6.108e+02 1.526e+03, threshold=9.466e+02, percent-clipped=4.0 +2023-02-05 21:45:23,371 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19636.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:45:42,898 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19663.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:45:45,373 INFO [train.py:901] (1/4) Epoch 3, batch 3500, loss[loss=0.2934, simple_loss=0.3443, pruned_loss=0.1213, over 7541.00 frames. ], tot_loss[loss=0.3381, simple_loss=0.3837, pruned_loss=0.1463, over 1612435.25 frames. ], batch size: 18, lr: 2.31e-02, grad_scale: 16.0 +2023-02-05 21:45:58,038 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-05 21:45:59,530 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19688.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:46:12,432 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 +2023-02-05 21:46:19,301 INFO [train.py:901] (1/4) Epoch 3, batch 3550, loss[loss=0.342, simple_loss=0.4014, pruned_loss=0.1413, over 8363.00 frames. ], tot_loss[loss=0.3391, simple_loss=0.3848, pruned_loss=0.1467, over 1611829.54 frames. ], batch size: 24, lr: 2.31e-02, grad_scale: 16.0 +2023-02-05 21:46:19,968 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 3.514e+02 4.193e+02 5.166e+02 1.109e+03, threshold=8.387e+02, percent-clipped=2.0 +2023-02-05 21:46:22,491 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.17 vs. limit=5.0 +2023-02-05 21:46:46,365 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19756.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:46:54,375 INFO [train.py:901] (1/4) Epoch 3, batch 3600, loss[loss=0.3664, simple_loss=0.4117, pruned_loss=0.1606, over 8759.00 frames. ], tot_loss[loss=0.3373, simple_loss=0.3836, pruned_loss=0.1455, over 1612761.76 frames. ], batch size: 30, lr: 2.31e-02, grad_scale: 16.0 +2023-02-05 21:47:04,332 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.58 vs. limit=5.0 +2023-02-05 21:47:17,891 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-05 21:47:28,230 INFO [train.py:901] (1/4) Epoch 3, batch 3650, loss[loss=0.2913, simple_loss=0.3209, pruned_loss=0.1309, over 6413.00 frames. ], tot_loss[loss=0.3357, simple_loss=0.3821, pruned_loss=0.1447, over 1610310.14 frames. ], batch size: 14, lr: 2.30e-02, grad_scale: 16.0 +2023-02-05 21:47:28,894 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 3.610e+02 4.497e+02 5.952e+02 1.837e+03, threshold=8.994e+02, percent-clipped=7.0 +2023-02-05 21:47:54,114 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-05 21:47:58,734 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-05 21:47:59,615 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19862.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:48:02,764 INFO [train.py:901] (1/4) Epoch 3, batch 3700, loss[loss=0.295, simple_loss=0.3611, pruned_loss=0.1145, over 8287.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3818, pruned_loss=0.1445, over 1611302.50 frames. ], batch size: 23, lr: 2.30e-02, grad_scale: 16.0 +2023-02-05 21:48:05,644 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19871.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:48:12,204 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3673, 1.5735, 1.3662, 1.1318, 1.6527, 1.5134, 1.5720, 1.9261], + device='cuda:1'), covar=tensor([0.0843, 0.1670, 0.2510, 0.1930, 0.0901, 0.2024, 0.1094, 0.0670], + device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0226, 0.0264, 0.0227, 0.0190, 0.0228, 0.0187, 0.0187], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0006, 0.0005, 0.0006, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 21:48:17,455 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19887.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:48:37,840 INFO [train.py:901] (1/4) Epoch 3, batch 3750, loss[loss=0.3822, simple_loss=0.4225, pruned_loss=0.171, over 8251.00 frames. ], tot_loss[loss=0.3361, simple_loss=0.3828, pruned_loss=0.1447, over 1613553.78 frames. ], batch size: 24, lr: 2.30e-02, grad_scale: 16.0 +2023-02-05 21:48:38,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 3.342e+02 4.116e+02 5.480e+02 1.463e+03, threshold=8.233e+02, percent-clipped=1.0 +2023-02-05 21:49:12,036 INFO [train.py:901] (1/4) Epoch 3, batch 3800, loss[loss=0.3361, simple_loss=0.3942, pruned_loss=0.1391, over 8467.00 frames. ], tot_loss[loss=0.3334, simple_loss=0.3807, pruned_loss=0.143, over 1613592.64 frames. ], batch size: 27, lr: 2.30e-02, grad_scale: 16.0 +2023-02-05 21:49:19,738 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-05 21:49:20,774 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19980.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:49:21,520 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6788, 1.2814, 5.7300, 1.8643, 4.9522, 4.8570, 5.2664, 5.2061], + device='cuda:1'), covar=tensor([0.0262, 0.3452, 0.0210, 0.1965, 0.0742, 0.0322, 0.0305, 0.0340], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0428, 0.0294, 0.0329, 0.0388, 0.0310, 0.0308, 0.0332], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 21:49:32,281 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4370, 1.9702, 3.1986, 0.9906, 2.1244, 1.7524, 1.6717, 1.6249], + device='cuda:1'), covar=tensor([0.1604, 0.1412, 0.0580, 0.2667, 0.1363, 0.2109, 0.1216, 0.2156], + device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0406, 0.0487, 0.0493, 0.0544, 0.0486, 0.0431, 0.0548], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:49:37,390 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-05 21:49:48,447 INFO [train.py:901] (1/4) Epoch 3, batch 3850, loss[loss=0.3547, simple_loss=0.4057, pruned_loss=0.1518, over 8608.00 frames. ], tot_loss[loss=0.3353, simple_loss=0.3822, pruned_loss=0.1442, over 1613023.52 frames. ], batch size: 34, lr: 2.29e-02, grad_scale: 16.0 +2023-02-05 21:49:49,085 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 3.536e+02 4.444e+02 5.257e+02 1.055e+03, threshold=8.889e+02, percent-clipped=4.0 +2023-02-05 21:49:50,510 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0483, 1.1585, 1.0663, 0.9127, 0.6964, 1.0980, 0.0349, 0.9978], + device='cuda:1'), covar=tensor([0.2380, 0.1241, 0.1204, 0.1701, 0.4260, 0.1111, 0.5220, 0.1652], + device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0093, 0.0080, 0.0144, 0.0167, 0.0076, 0.0150, 0.0107], + device='cuda:1'), out_proj_covar=tensor([1.3875e-04, 1.2032e-04, 1.0436e-04, 1.7446e-04, 1.9680e-04, 9.9052e-05, + 1.8260e-04, 1.4189e-04], device='cuda:1') +2023-02-05 21:50:01,939 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-05 21:50:02,725 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20038.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:50:11,369 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1683, 1.6994, 3.1778, 1.5704, 2.3872, 3.6120, 3.3141, 3.1730], + device='cuda:1'), covar=tensor([0.0920, 0.1238, 0.0451, 0.1609, 0.0685, 0.0173, 0.0308, 0.0482], + device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0256, 0.0195, 0.0256, 0.0201, 0.0168, 0.0169, 0.0242], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 21:50:22,591 INFO [train.py:901] (1/4) Epoch 3, batch 3900, loss[loss=0.3183, simple_loss=0.3652, pruned_loss=0.1357, over 8193.00 frames. ], tot_loss[loss=0.3384, simple_loss=0.3845, pruned_loss=0.1461, over 1615051.66 frames. ], batch size: 23, lr: 2.29e-02, grad_scale: 16.0 +2023-02-05 21:50:22,832 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1073, 1.8537, 3.5055, 2.9994, 2.6234, 1.5280, 1.1405, 1.3943], + device='cuda:1'), covar=tensor([0.1702, 0.1662, 0.0277, 0.0497, 0.0665, 0.1077, 0.1187, 0.1391], + device='cuda:1'), in_proj_covar=tensor([0.0588, 0.0504, 0.0417, 0.0463, 0.0564, 0.0476, 0.0498, 0.0491], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 21:50:41,539 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20095.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:50:42,863 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.5527, 1.4556, 5.5861, 2.2457, 4.9371, 4.6682, 5.0077, 5.1751], + device='cuda:1'), covar=tensor([0.0228, 0.3142, 0.0164, 0.1527, 0.0649, 0.0312, 0.0270, 0.0244], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0421, 0.0296, 0.0326, 0.0388, 0.0306, 0.0306, 0.0329], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 21:50:50,426 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20107.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:50:56,844 INFO [train.py:901] (1/4) Epoch 3, batch 3950, loss[loss=0.2993, simple_loss=0.3667, pruned_loss=0.116, over 8362.00 frames. ], tot_loss[loss=0.337, simple_loss=0.3836, pruned_loss=0.1452, over 1616041.49 frames. ], batch size: 24, lr: 2.29e-02, grad_scale: 16.0 +2023-02-05 21:50:57,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.523e+02 3.492e+02 4.461e+02 6.032e+02 1.371e+03, threshold=8.922e+02, percent-clipped=4.0 +2023-02-05 21:51:05,055 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20127.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:51:06,984 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20130.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:51:21,462 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20152.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:51:31,210 INFO [train.py:901] (1/4) Epoch 3, batch 4000, loss[loss=0.3479, simple_loss=0.4044, pruned_loss=0.1457, over 8308.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3825, pruned_loss=0.1441, over 1616574.53 frames. ], batch size: 23, lr: 2.29e-02, grad_scale: 16.0 +2023-02-05 21:51:38,888 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.9114, 1.1708, 4.1478, 1.6502, 3.4153, 3.3994, 3.6319, 3.6484], + device='cuda:1'), covar=tensor([0.0415, 0.3087, 0.0290, 0.1774, 0.1015, 0.0475, 0.0409, 0.0456], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0410, 0.0291, 0.0321, 0.0391, 0.0303, 0.0304, 0.0329], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-05 21:51:50,790 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7072, 1.9442, 3.6485, 1.0216, 2.5138, 1.8900, 1.5896, 1.8863], + device='cuda:1'), covar=tensor([0.1190, 0.1593, 0.0385, 0.2442, 0.1186, 0.1908, 0.1136, 0.2097], + device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0400, 0.0463, 0.0477, 0.0532, 0.0476, 0.0419, 0.0533], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:52:05,183 INFO [train.py:901] (1/4) Epoch 3, batch 4050, loss[loss=0.3333, simple_loss=0.3659, pruned_loss=0.1503, over 7227.00 frames. ], tot_loss[loss=0.3338, simple_loss=0.3812, pruned_loss=0.1432, over 1615515.89 frames. ], batch size: 16, lr: 2.28e-02, grad_scale: 16.0 +2023-02-05 21:52:05,854 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.226e+02 3.505e+02 4.242e+02 5.307e+02 1.364e+03, threshold=8.485e+02, percent-clipped=4.0 +2023-02-05 21:52:09,348 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20222.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:52:37,132 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5183, 1.8421, 2.8433, 1.1279, 2.3259, 1.7271, 1.5480, 1.7112], + device='cuda:1'), covar=tensor([0.1255, 0.1358, 0.0473, 0.2277, 0.0799, 0.1881, 0.1105, 0.1558], + device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0402, 0.0472, 0.0484, 0.0531, 0.0481, 0.0423, 0.0536], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:52:40,359 INFO [train.py:901] (1/4) Epoch 3, batch 4100, loss[loss=0.376, simple_loss=0.4104, pruned_loss=0.1708, over 6904.00 frames. ], tot_loss[loss=0.3353, simple_loss=0.3827, pruned_loss=0.1439, over 1615079.53 frames. ], batch size: 71, lr: 2.28e-02, grad_scale: 8.0 +2023-02-05 21:52:41,926 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0420, 2.1960, 1.6865, 2.8499, 1.6073, 1.3579, 1.7728, 2.3857], + device='cuda:1'), covar=tensor([0.1140, 0.1633, 0.2021, 0.0587, 0.2001, 0.2690, 0.2207, 0.1204], + device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0312, 0.0308, 0.0229, 0.0296, 0.0312, 0.0335, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 21:52:57,065 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-05 21:53:14,414 INFO [train.py:901] (1/4) Epoch 3, batch 4150, loss[loss=0.3342, simple_loss=0.3735, pruned_loss=0.1474, over 8601.00 frames. ], tot_loss[loss=0.3361, simple_loss=0.3835, pruned_loss=0.1443, over 1620912.37 frames. ], batch size: 39, lr: 2.28e-02, grad_scale: 8.0 +2023-02-05 21:53:15,797 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.167e+02 3.849e+02 4.660e+02 5.932e+02 1.097e+03, threshold=9.320e+02, percent-clipped=6.0 +2023-02-05 21:53:38,380 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20351.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:53:49,993 INFO [train.py:901] (1/4) Epoch 3, batch 4200, loss[loss=0.2892, simple_loss=0.349, pruned_loss=0.1147, over 7665.00 frames. ], tot_loss[loss=0.3359, simple_loss=0.383, pruned_loss=0.1444, over 1619526.84 frames. ], batch size: 19, lr: 2.27e-02, grad_scale: 8.0 +2023-02-05 21:53:55,427 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-05 21:53:56,308 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:00,121 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20382.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:16,277 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20406.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:16,812 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-05 21:54:24,064 INFO [train.py:901] (1/4) Epoch 3, batch 4250, loss[loss=0.3688, simple_loss=0.4153, pruned_loss=0.1611, over 8460.00 frames. ], tot_loss[loss=0.3359, simple_loss=0.3829, pruned_loss=0.1444, over 1618022.22 frames. ], batch size: 27, lr: 2.27e-02, grad_scale: 8.0 +2023-02-05 21:54:25,369 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.360e+02 3.627e+02 5.036e+02 6.332e+02 1.636e+03, threshold=1.007e+03, percent-clipped=4.0 +2023-02-05 21:54:29,691 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5381, 2.0196, 3.3016, 1.1408, 2.3357, 1.7728, 1.5221, 1.8937], + device='cuda:1'), covar=tensor([0.1123, 0.1159, 0.0453, 0.2164, 0.0906, 0.1682, 0.1034, 0.1548], + device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0398, 0.0471, 0.0485, 0.0532, 0.0473, 0.0419, 0.0533], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:54:31,104 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-05 21:54:37,015 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20436.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:47,726 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20451.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:59,264 INFO [train.py:901] (1/4) Epoch 3, batch 4300, loss[loss=0.3378, simple_loss=0.3781, pruned_loss=0.1488, over 7685.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3822, pruned_loss=0.1443, over 1611828.80 frames. ], batch size: 18, lr: 2.27e-02, grad_scale: 8.0 +2023-02-05 21:55:04,855 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20474.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:55:20,187 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20497.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:55:33,536 INFO [train.py:901] (1/4) Epoch 3, batch 4350, loss[loss=0.3134, simple_loss=0.3545, pruned_loss=0.1361, over 7810.00 frames. ], tot_loss[loss=0.335, simple_loss=0.3819, pruned_loss=0.1441, over 1617391.89 frames. ], batch size: 20, lr: 2.27e-02, grad_scale: 8.0 +2023-02-05 21:55:34,898 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 3.452e+02 4.356e+02 5.638e+02 1.577e+03, threshold=8.711e+02, percent-clipped=2.0 +2023-02-05 21:55:46,534 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-05 21:55:55,316 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4889, 2.0517, 3.6037, 1.1454, 2.4896, 1.7866, 1.5476, 1.9710], + device='cuda:1'), covar=tensor([0.1197, 0.1265, 0.0424, 0.2169, 0.1021, 0.1807, 0.1059, 0.1862], + device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0406, 0.0478, 0.0494, 0.0534, 0.0475, 0.0426, 0.0538], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:55:57,857 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4972, 2.0741, 1.2971, 1.6979, 1.7536, 1.2308, 1.4686, 1.9950], + device='cuda:1'), covar=tensor([0.0938, 0.0291, 0.0950, 0.0579, 0.0578, 0.1003, 0.0967, 0.0571], + device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0241, 0.0336, 0.0319, 0.0339, 0.0321, 0.0363, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 21:56:06,881 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20566.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:56:06,990 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20566.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:56:07,507 INFO [train.py:901] (1/4) Epoch 3, batch 4400, loss[loss=0.3091, simple_loss=0.3476, pruned_loss=0.1353, over 6812.00 frames. ], tot_loss[loss=0.3374, simple_loss=0.3833, pruned_loss=0.1458, over 1613718.48 frames. ], batch size: 15, lr: 2.26e-02, grad_scale: 8.0 +2023-02-05 21:56:11,290 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-05 21:56:24,123 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20589.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:56:27,484 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-05 21:56:36,782 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20606.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:56:44,139 INFO [train.py:901] (1/4) Epoch 3, batch 4450, loss[loss=0.3712, simple_loss=0.4197, pruned_loss=0.1613, over 8615.00 frames. ], tot_loss[loss=0.3359, simple_loss=0.383, pruned_loss=0.1444, over 1614691.32 frames. ], batch size: 31, lr: 2.26e-02, grad_scale: 8.0 +2023-02-05 21:56:45,438 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 3.404e+02 4.420e+02 6.069e+02 1.310e+03, threshold=8.839e+02, percent-clipped=8.0 +2023-02-05 21:57:07,990 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3207, 1.6488, 1.3921, 1.2837, 1.6593, 1.4754, 1.6310, 1.6843], + device='cuda:1'), covar=tensor([0.0732, 0.1377, 0.2075, 0.1643, 0.0795, 0.1705, 0.0992, 0.0686], + device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0218, 0.0256, 0.0220, 0.0185, 0.0220, 0.0181, 0.0182], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0006, 0.0005, 0.0006, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 21:57:12,691 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7488, 1.2511, 3.2408, 1.4510, 2.2110, 3.5020, 3.2791, 3.0834], + device='cuda:1'), covar=tensor([0.1048, 0.1691, 0.0337, 0.1936, 0.0724, 0.0267, 0.0349, 0.0518], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0270, 0.0208, 0.0269, 0.0213, 0.0180, 0.0183, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:57:18,488 INFO [train.py:901] (1/4) Epoch 3, batch 4500, loss[loss=0.3422, simple_loss=0.3908, pruned_loss=0.1468, over 8658.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3823, pruned_loss=0.1443, over 1609586.09 frames. ], batch size: 34, lr: 2.26e-02, grad_scale: 8.0 +2023-02-05 21:57:20,855 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-05 21:57:21,680 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6269, 2.5553, 1.5211, 1.9480, 1.9348, 1.4746, 1.8692, 2.1615], + device='cuda:1'), covar=tensor([0.1071, 0.0342, 0.0893, 0.0493, 0.0522, 0.0976, 0.0751, 0.0671], + device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0241, 0.0329, 0.0315, 0.0342, 0.0320, 0.0357, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 21:57:27,450 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20681.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:57:30,258 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.01 vs. limit=5.0 +2023-02-05 21:57:53,171 INFO [train.py:901] (1/4) Epoch 3, batch 4550, loss[loss=0.3159, simple_loss=0.359, pruned_loss=0.1364, over 7540.00 frames. ], tot_loss[loss=0.3351, simple_loss=0.3822, pruned_loss=0.144, over 1607476.11 frames. ], batch size: 18, lr: 2.26e-02, grad_scale: 8.0 +2023-02-05 21:57:54,484 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.207e+02 3.483e+02 4.570e+02 6.300e+02 1.347e+03, threshold=9.139e+02, percent-clipped=2.0 +2023-02-05 21:57:54,738 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6576, 2.2055, 4.5172, 0.9793, 2.7558, 2.2893, 1.5266, 2.3240], + device='cuda:1'), covar=tensor([0.1365, 0.1688, 0.0530, 0.2887, 0.1174, 0.1957, 0.1237, 0.2234], + device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0411, 0.0479, 0.0497, 0.0548, 0.0484, 0.0425, 0.0550], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 21:58:14,826 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20750.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:58:17,073 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20753.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:58:26,235 INFO [train.py:901] (1/4) Epoch 3, batch 4600, loss[loss=0.3169, simple_loss=0.3676, pruned_loss=0.1332, over 7706.00 frames. ], tot_loss[loss=0.3355, simple_loss=0.382, pruned_loss=0.1445, over 1611552.82 frames. ], batch size: 18, lr: 2.25e-02, grad_scale: 8.0 +2023-02-05 21:58:34,567 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20778.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:58:35,792 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20780.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:00,491 INFO [train.py:901] (1/4) Epoch 3, batch 4650, loss[loss=0.2867, simple_loss=0.3496, pruned_loss=0.1119, over 7976.00 frames. ], tot_loss[loss=0.3363, simple_loss=0.3823, pruned_loss=0.1451, over 1607618.71 frames. ], batch size: 21, lr: 2.25e-02, grad_scale: 8.0 +2023-02-05 21:59:02,525 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.001e+02 3.299e+02 4.239e+02 5.426e+02 9.400e+02, threshold=8.478e+02, percent-clipped=1.0 +2023-02-05 21:59:04,816 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20822.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:21,394 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20845.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:22,741 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20847.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:34,619 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20865.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:35,785 INFO [train.py:901] (1/4) Epoch 3, batch 4700, loss[loss=0.3464, simple_loss=0.4015, pruned_loss=0.1456, over 8448.00 frames. ], tot_loss[loss=0.3356, simple_loss=0.3822, pruned_loss=0.1445, over 1614936.92 frames. ], batch size: 27, lr: 2.25e-02, grad_scale: 8.0 +2023-02-05 21:59:37,914 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20870.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:54,881 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20895.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:00:08,976 INFO [train.py:901] (1/4) Epoch 3, batch 4750, loss[loss=0.3329, simple_loss=0.385, pruned_loss=0.1404, over 8321.00 frames. ], tot_loss[loss=0.3345, simple_loss=0.3812, pruned_loss=0.1439, over 1611062.60 frames. ], batch size: 25, lr: 2.25e-02, grad_scale: 8.0 +2023-02-05 22:00:10,302 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.464e+02 3.634e+02 4.432e+02 5.821e+02 1.296e+03, threshold=8.863e+02, percent-clipped=5.0 +2023-02-05 22:00:23,256 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20937.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:00:24,395 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-05 22:00:24,548 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6840, 1.4752, 3.3080, 1.1864, 2.2232, 3.6643, 3.4298, 3.1541], + device='cuda:1'), covar=tensor([0.1077, 0.1265, 0.0291, 0.1808, 0.0616, 0.0223, 0.0308, 0.0528], + device='cuda:1'), in_proj_covar=tensor([0.0227, 0.0255, 0.0197, 0.0255, 0.0204, 0.0174, 0.0175, 0.0250], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:00:26,473 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-05 22:00:32,640 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20950.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:00:41,442 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20962.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:00:44,560 INFO [train.py:901] (1/4) Epoch 3, batch 4800, loss[loss=0.3724, simple_loss=0.4149, pruned_loss=0.165, over 8340.00 frames. ], tot_loss[loss=0.3337, simple_loss=0.3808, pruned_loss=0.1432, over 1610258.11 frames. ], batch size: 25, lr: 2.24e-02, grad_scale: 8.0 +2023-02-05 22:01:18,124 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-05 22:01:18,796 INFO [train.py:901] (1/4) Epoch 3, batch 4850, loss[loss=0.3502, simple_loss=0.3825, pruned_loss=0.1589, over 7434.00 frames. ], tot_loss[loss=0.3321, simple_loss=0.3796, pruned_loss=0.1423, over 1607061.03 frames. ], batch size: 17, lr: 2.24e-02, grad_scale: 8.0 +2023-02-05 22:01:20,191 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.459e+02 3.687e+02 4.412e+02 5.668e+02 1.155e+03, threshold=8.825e+02, percent-clipped=6.0 +2023-02-05 22:01:37,496 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3910, 1.9499, 3.1735, 1.0158, 2.1897, 1.6455, 1.4781, 1.7705], + device='cuda:1'), covar=tensor([0.1283, 0.1330, 0.0480, 0.2501, 0.1148, 0.1849, 0.1162, 0.1815], + device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0410, 0.0484, 0.0501, 0.0551, 0.0475, 0.0432, 0.0547], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 22:01:42,459 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-02-05 22:01:51,977 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21065.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:01:53,054 INFO [train.py:901] (1/4) Epoch 3, batch 4900, loss[loss=0.3242, simple_loss=0.3548, pruned_loss=0.1468, over 7637.00 frames. ], tot_loss[loss=0.3326, simple_loss=0.3801, pruned_loss=0.1425, over 1609938.98 frames. ], batch size: 19, lr: 2.24e-02, grad_scale: 8.0 +2023-02-05 22:01:55,916 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21070.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:02:27,700 INFO [train.py:901] (1/4) Epoch 3, batch 4950, loss[loss=0.3417, simple_loss=0.3874, pruned_loss=0.148, over 8730.00 frames. ], tot_loss[loss=0.3331, simple_loss=0.3806, pruned_loss=0.1428, over 1608252.66 frames. ], batch size: 34, lr: 2.24e-02, grad_scale: 8.0 +2023-02-05 22:02:29,085 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.083e+02 3.569e+02 4.502e+02 6.229e+02 1.133e+03, threshold=9.004e+02, percent-clipped=2.0 +2023-02-05 22:02:30,676 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21121.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:02:41,332 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21136.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:02:48,073 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21146.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:02:51,488 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21151.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:02:53,464 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2074, 1.6885, 2.0342, 1.5464, 0.8518, 1.7672, 0.2681, 1.2170], + device='cuda:1'), covar=tensor([0.3447, 0.1709, 0.0786, 0.1873, 0.4412, 0.1205, 0.4778, 0.2230], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0106, 0.0082, 0.0152, 0.0171, 0.0081, 0.0149, 0.0115], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:03:01,832 INFO [train.py:901] (1/4) Epoch 3, batch 5000, loss[loss=0.2863, simple_loss=0.3353, pruned_loss=0.1187, over 7424.00 frames. ], tot_loss[loss=0.3314, simple_loss=0.3795, pruned_loss=0.1417, over 1611197.11 frames. ], batch size: 17, lr: 2.23e-02, grad_scale: 8.0 +2023-02-05 22:03:08,655 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21176.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:03:34,153 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0956, 1.4331, 1.5096, 1.1294, 1.5240, 1.4486, 1.4640, 1.5415], + device='cuda:1'), covar=tensor([0.0727, 0.1407, 0.1838, 0.1660, 0.0681, 0.1580, 0.0932, 0.0650], + device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0218, 0.0254, 0.0217, 0.0178, 0.0219, 0.0181, 0.0180], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0006, 0.0005, 0.0006, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 22:03:37,220 INFO [train.py:901] (1/4) Epoch 3, batch 5050, loss[loss=0.2858, simple_loss=0.3458, pruned_loss=0.1129, over 8204.00 frames. ], tot_loss[loss=0.3299, simple_loss=0.378, pruned_loss=0.1409, over 1607337.83 frames. ], batch size: 23, lr: 2.23e-02, grad_scale: 8.0 +2023-02-05 22:03:38,541 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 3.325e+02 4.224e+02 5.254e+02 1.187e+03, threshold=8.447e+02, percent-clipped=3.0 +2023-02-05 22:03:57,070 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-05 22:04:11,239 INFO [train.py:901] (1/4) Epoch 3, batch 5100, loss[loss=0.2896, simple_loss=0.3422, pruned_loss=0.1185, over 7818.00 frames. ], tot_loss[loss=0.33, simple_loss=0.378, pruned_loss=0.141, over 1607877.45 frames. ], batch size: 20, lr: 2.23e-02, grad_scale: 8.0 +2023-02-05 22:04:28,234 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4702, 1.8796, 1.9788, 0.7704, 1.9827, 1.3514, 0.5154, 1.6575], + device='cuda:1'), covar=tensor([0.0130, 0.0072, 0.0096, 0.0151, 0.0086, 0.0248, 0.0224, 0.0067], + device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0161, 0.0135, 0.0208, 0.0159, 0.0276, 0.0219, 0.0189], + device='cuda:1'), out_proj_covar=tensor([1.0640e-04, 7.5361e-05, 6.1075e-05, 9.3885e-05, 7.5659e-05, 1.3855e-04, + 1.0545e-04, 8.7761e-05], device='cuda:1') +2023-02-05 22:04:46,363 INFO [train.py:901] (1/4) Epoch 3, batch 5150, loss[loss=0.4568, simple_loss=0.463, pruned_loss=0.2253, over 7393.00 frames. ], tot_loss[loss=0.3287, simple_loss=0.3772, pruned_loss=0.1401, over 1608057.73 frames. ], batch size: 71, lr: 2.23e-02, grad_scale: 8.0 +2023-02-05 22:04:47,674 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.142e+02 3.453e+02 4.061e+02 5.332e+02 1.278e+03, threshold=8.122e+02, percent-clipped=4.0 +2023-02-05 22:04:50,019 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21321.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:04:56,026 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21330.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:05:06,483 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21346.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:05:06,543 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21346.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:05:19,999 INFO [train.py:901] (1/4) Epoch 3, batch 5200, loss[loss=0.3589, simple_loss=0.3814, pruned_loss=0.1682, over 8247.00 frames. ], tot_loss[loss=0.3292, simple_loss=0.3774, pruned_loss=0.1405, over 1608426.69 frames. ], batch size: 22, lr: 2.22e-02, grad_scale: 8.0 +2023-02-05 22:05:43,302 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5589, 2.5793, 1.5993, 1.9358, 1.8880, 1.1226, 1.8142, 2.0318], + device='cuda:1'), covar=tensor([0.1383, 0.0401, 0.1108, 0.0777, 0.0955, 0.1578, 0.1164, 0.0829], + device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0235, 0.0325, 0.0316, 0.0338, 0.0321, 0.0347, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 22:05:52,822 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21414.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:05:54,863 INFO [train.py:901] (1/4) Epoch 3, batch 5250, loss[loss=0.3301, simple_loss=0.3924, pruned_loss=0.1339, over 8525.00 frames. ], tot_loss[loss=0.3315, simple_loss=0.3787, pruned_loss=0.1421, over 1606011.34 frames. ], batch size: 28, lr: 2.22e-02, grad_scale: 8.0 +2023-02-05 22:05:54,880 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-05 22:05:56,257 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 3.353e+02 4.281e+02 5.765e+02 2.364e+03, threshold=8.563e+02, percent-clipped=11.0 +2023-02-05 22:06:15,346 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1777, 4.2031, 3.8835, 1.7270, 3.6898, 3.6613, 4.0167, 3.0769], + device='cuda:1'), covar=tensor([0.1028, 0.0628, 0.0928, 0.4668, 0.0645, 0.0711, 0.1335, 0.0876], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0234, 0.0284, 0.0367, 0.0264, 0.0211, 0.0263, 0.0202], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 22:06:30,410 INFO [train.py:901] (1/4) Epoch 3, batch 5300, loss[loss=0.3884, simple_loss=0.4142, pruned_loss=0.1813, over 8664.00 frames. ], tot_loss[loss=0.3316, simple_loss=0.3793, pruned_loss=0.1419, over 1609283.24 frames. ], batch size: 39, lr: 2.22e-02, grad_scale: 8.0 +2023-02-05 22:06:39,463 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21480.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:06:57,980 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-02-05 22:07:04,804 INFO [train.py:901] (1/4) Epoch 3, batch 5350, loss[loss=0.3074, simple_loss=0.3574, pruned_loss=0.1287, over 7801.00 frames. ], tot_loss[loss=0.332, simple_loss=0.3801, pruned_loss=0.142, over 1613305.51 frames. ], batch size: 20, lr: 2.22e-02, grad_scale: 8.0 +2023-02-05 22:07:06,080 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.137e+02 3.338e+02 4.128e+02 5.460e+02 1.129e+03, threshold=8.255e+02, percent-clipped=3.0 +2023-02-05 22:07:13,625 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21529.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:07:40,175 INFO [train.py:901] (1/4) Epoch 3, batch 5400, loss[loss=0.3578, simple_loss=0.3962, pruned_loss=0.1597, over 7803.00 frames. ], tot_loss[loss=0.3349, simple_loss=0.3822, pruned_loss=0.1438, over 1614244.26 frames. ], batch size: 20, lr: 2.21e-02, grad_scale: 8.0 +2023-02-05 22:07:52,667 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-05 22:07:59,350 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:08:12,087 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:08:14,749 INFO [train.py:901] (1/4) Epoch 3, batch 5450, loss[loss=0.3217, simple_loss=0.3807, pruned_loss=0.1313, over 8481.00 frames. ], tot_loss[loss=0.3351, simple_loss=0.3826, pruned_loss=0.1438, over 1614382.96 frames. ], batch size: 25, lr: 2.21e-02, grad_scale: 8.0 +2023-02-05 22:08:16,068 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.864e+02 3.746e+02 4.366e+02 5.874e+02 2.172e+03, threshold=8.732e+02, percent-clipped=6.0 +2023-02-05 22:08:19,189 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-05 22:08:24,284 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21631.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:08:41,812 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-05 22:08:49,813 INFO [train.py:901] (1/4) Epoch 3, batch 5500, loss[loss=0.2579, simple_loss=0.3087, pruned_loss=0.1036, over 7410.00 frames. ], tot_loss[loss=0.3328, simple_loss=0.3811, pruned_loss=0.1423, over 1614771.11 frames. ], batch size: 17, lr: 2.21e-02, grad_scale: 8.0 +2023-02-05 22:08:55,224 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21674.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:09:05,892 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21690.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:09:23,546 INFO [train.py:901] (1/4) Epoch 3, batch 5550, loss[loss=0.3151, simple_loss=0.3624, pruned_loss=0.1338, over 7656.00 frames. ], tot_loss[loss=0.3313, simple_loss=0.3804, pruned_loss=0.1411, over 1614406.73 frames. ], batch size: 19, lr: 2.21e-02, grad_scale: 8.0 +2023-02-05 22:09:24,912 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.889e+02 3.296e+02 4.063e+02 5.206e+02 8.291e+02, threshold=8.125e+02, percent-clipped=0.0 +2023-02-05 22:09:28,458 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3042, 1.2785, 2.7799, 1.1482, 2.0636, 2.9175, 2.6145, 2.5219], + device='cuda:1'), covar=tensor([0.1384, 0.1533, 0.0441, 0.2038, 0.0665, 0.0370, 0.0691, 0.0704], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0261, 0.0200, 0.0254, 0.0201, 0.0180, 0.0177, 0.0252], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:09:58,432 INFO [train.py:901] (1/4) Epoch 3, batch 5600, loss[loss=0.4309, simple_loss=0.4508, pruned_loss=0.2055, over 8365.00 frames. ], tot_loss[loss=0.3304, simple_loss=0.3796, pruned_loss=0.1406, over 1609224.40 frames. ], batch size: 24, lr: 2.20e-02, grad_scale: 8.0 +2023-02-05 22:10:08,490 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21781.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:11,832 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21785.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:14,466 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21789.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:25,386 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21805.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:28,825 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21810.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:33,432 INFO [train.py:901] (1/4) Epoch 3, batch 5650, loss[loss=0.2764, simple_loss=0.3358, pruned_loss=0.1085, over 8143.00 frames. ], tot_loss[loss=0.328, simple_loss=0.3774, pruned_loss=0.1392, over 1606761.06 frames. ], batch size: 22, lr: 2.20e-02, grad_scale: 8.0 +2023-02-05 22:10:34,802 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.984e+02 3.614e+02 4.526e+02 5.980e+02 8.654e+02, threshold=9.051e+02, percent-clipped=4.0 +2023-02-05 22:10:45,286 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-05 22:10:56,855 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21851.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:58,914 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4993, 2.0581, 3.4134, 2.8174, 2.7163, 2.0882, 1.5102, 1.5600], + device='cuda:1'), covar=tensor([0.1024, 0.1210, 0.0202, 0.0488, 0.0526, 0.0512, 0.0713, 0.1125], + device='cuda:1'), in_proj_covar=tensor([0.0627, 0.0541, 0.0457, 0.0499, 0.0610, 0.0502, 0.0518, 0.0523], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:11:07,114 INFO [train.py:901] (1/4) Epoch 3, batch 5700, loss[loss=0.3546, simple_loss=0.3897, pruned_loss=0.1598, over 8353.00 frames. ], tot_loss[loss=0.3309, simple_loss=0.379, pruned_loss=0.1414, over 1603326.52 frames. ], batch size: 26, lr: 2.20e-02, grad_scale: 8.0 +2023-02-05 22:11:07,936 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21868.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:11:13,371 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21876.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:11:42,901 INFO [train.py:901] (1/4) Epoch 3, batch 5750, loss[loss=0.3923, simple_loss=0.4165, pruned_loss=0.184, over 8471.00 frames. ], tot_loss[loss=0.3295, simple_loss=0.3777, pruned_loss=0.1406, over 1604670.56 frames. ], batch size: 25, lr: 2.20e-02, grad_scale: 8.0 +2023-02-05 22:11:44,219 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.339e+02 3.657e+02 4.422e+02 5.345e+02 1.248e+03, threshold=8.845e+02, percent-clipped=3.0 +2023-02-05 22:11:49,709 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-05 22:11:53,992 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5802, 2.2623, 4.5512, 1.0674, 2.9220, 2.1347, 1.7622, 2.5585], + device='cuda:1'), covar=tensor([0.1336, 0.1690, 0.0488, 0.2789, 0.1155, 0.2106, 0.1176, 0.1888], + device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0416, 0.0490, 0.0504, 0.0545, 0.0485, 0.0425, 0.0549], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 22:12:02,091 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-02-05 22:12:10,247 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21957.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:12:16,968 INFO [train.py:901] (1/4) Epoch 3, batch 5800, loss[loss=0.3574, simple_loss=0.3919, pruned_loss=0.1615, over 7970.00 frames. ], tot_loss[loss=0.3292, simple_loss=0.3771, pruned_loss=0.1406, over 1603402.63 frames. ], batch size: 21, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:12:22,434 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21975.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:12:31,003 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21988.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:12:52,153 INFO [train.py:901] (1/4) Epoch 3, batch 5850, loss[loss=0.317, simple_loss=0.3741, pruned_loss=0.1299, over 8337.00 frames. ], tot_loss[loss=0.3271, simple_loss=0.3754, pruned_loss=0.1394, over 1602388.49 frames. ], batch size: 26, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:12:53,399 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.077e+02 3.662e+02 4.461e+02 5.594e+02 1.608e+03, threshold=8.923e+02, percent-clipped=8.0 +2023-02-05 22:12:55,624 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9509, 3.7796, 2.3623, 2.5414, 2.9053, 1.7439, 2.2468, 2.5485], + device='cuda:1'), covar=tensor([0.1675, 0.0528, 0.0988, 0.1066, 0.0721, 0.1324, 0.1410, 0.1178], + device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0249, 0.0327, 0.0316, 0.0340, 0.0323, 0.0357, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 22:13:11,807 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22045.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:13:22,197 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22061.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:13:25,935 INFO [train.py:901] (1/4) Epoch 3, batch 5900, loss[loss=0.3387, simple_loss=0.3652, pruned_loss=0.1561, over 7557.00 frames. ], tot_loss[loss=0.3266, simple_loss=0.3749, pruned_loss=0.1391, over 1605293.50 frames. ], batch size: 18, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:13:28,813 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22070.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:13:30,188 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22072.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:13:39,580 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22086.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:13:42,233 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22090.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:14:00,344 INFO [train.py:901] (1/4) Epoch 3, batch 5950, loss[loss=0.2669, simple_loss=0.3305, pruned_loss=0.1016, over 8076.00 frames. ], tot_loss[loss=0.3287, simple_loss=0.3766, pruned_loss=0.1404, over 1608931.23 frames. ], batch size: 21, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:14:02,408 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.045e+02 3.353e+02 4.485e+02 5.691e+02 1.558e+03, threshold=8.970e+02, percent-clipped=6.0 +2023-02-05 22:14:07,175 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22125.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:14:23,952 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22148.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:14:36,958 INFO [train.py:901] (1/4) Epoch 3, batch 6000, loss[loss=0.343, simple_loss=0.3814, pruned_loss=0.1523, over 7918.00 frames. ], tot_loss[loss=0.3271, simple_loss=0.3754, pruned_loss=0.1394, over 1606173.01 frames. ], batch size: 20, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:14:36,958 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 22:14:49,936 INFO [train.py:935] (1/4) Epoch 3, validation: loss=0.2472, simple_loss=0.3383, pruned_loss=0.07805, over 944034.00 frames. +2023-02-05 22:14:49,936 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-05 22:15:08,345 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22194.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:21,644 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22212.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:25,115 INFO [train.py:901] (1/4) Epoch 3, batch 6050, loss[loss=0.3827, simple_loss=0.4084, pruned_loss=0.1785, over 7935.00 frames. ], tot_loss[loss=0.3258, simple_loss=0.3745, pruned_loss=0.1386, over 1609195.42 frames. ], batch size: 20, lr: 2.18e-02, grad_scale: 8.0 +2023-02-05 22:15:26,475 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.554e+02 3.417e+02 4.364e+02 5.364e+02 3.571e+03, threshold=8.727e+02, percent-clipped=6.0 +2023-02-05 22:15:36,164 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22233.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:40,933 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22240.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:51,050 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22255.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:59,767 INFO [train.py:901] (1/4) Epoch 3, batch 6100, loss[loss=0.2872, simple_loss=0.355, pruned_loss=0.1096, over 8095.00 frames. ], tot_loss[loss=0.328, simple_loss=0.3768, pruned_loss=0.1397, over 1607431.94 frames. ], batch size: 23, lr: 2.18e-02, grad_scale: 16.0 +2023-02-05 22:16:12,751 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7759, 4.1472, 2.5273, 2.6872, 2.9347, 2.0202, 2.4142, 2.8249], + device='cuda:1'), covar=tensor([0.1371, 0.0235, 0.0675, 0.0740, 0.0536, 0.0994, 0.0987, 0.0754], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0238, 0.0320, 0.0305, 0.0327, 0.0309, 0.0343, 0.0300], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 22:16:18,458 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-05 22:16:35,122 INFO [train.py:901] (1/4) Epoch 3, batch 6150, loss[loss=0.306, simple_loss=0.3712, pruned_loss=0.1204, over 8482.00 frames. ], tot_loss[loss=0.3295, simple_loss=0.3784, pruned_loss=0.1402, over 1610625.17 frames. ], batch size: 27, lr: 2.18e-02, grad_scale: 16.0 +2023-02-05 22:16:36,462 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.469e+02 3.615e+02 4.380e+02 5.688e+02 1.525e+03, threshold=8.759e+02, percent-clipped=2.0 +2023-02-05 22:16:41,831 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22327.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:16:42,483 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22328.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:16:45,082 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22332.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:16:50,862 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-05 22:16:51,479 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-05 22:16:54,536 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22346.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:16:59,259 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22353.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:17:08,709 INFO [train.py:901] (1/4) Epoch 3, batch 6200, loss[loss=0.3291, simple_loss=0.3771, pruned_loss=0.1406, over 8198.00 frames. ], tot_loss[loss=0.3282, simple_loss=0.3774, pruned_loss=0.1396, over 1611396.64 frames. ], batch size: 23, lr: 2.18e-02, grad_scale: 16.0 +2023-02-05 22:17:11,709 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22371.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:17:14,321 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7222, 3.6500, 3.2585, 1.9539, 3.2322, 3.1727, 3.4420, 2.7931], + device='cuda:1'), covar=tensor([0.0991, 0.0719, 0.1007, 0.4254, 0.0783, 0.0947, 0.1211, 0.0931], + device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0250, 0.0281, 0.0375, 0.0265, 0.0217, 0.0263, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 22:17:34,075 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 +2023-02-05 22:17:34,524 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22403.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:17:44,420 INFO [train.py:901] (1/4) Epoch 3, batch 6250, loss[loss=0.2837, simple_loss=0.3537, pruned_loss=0.1068, over 8456.00 frames. ], tot_loss[loss=0.3296, simple_loss=0.3782, pruned_loss=0.1405, over 1611868.93 frames. ], batch size: 29, lr: 2.17e-02, grad_scale: 16.0 +2023-02-05 22:17:45,753 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.208e+02 3.506e+02 4.308e+02 5.585e+02 1.214e+03, threshold=8.617e+02, percent-clipped=6.0 +2023-02-05 22:17:57,836 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-05 22:18:04,325 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2686, 1.2594, 2.2518, 1.1326, 2.1114, 2.4155, 2.3776, 2.0942], + device='cuda:1'), covar=tensor([0.1045, 0.1150, 0.0435, 0.1884, 0.0452, 0.0362, 0.0371, 0.0712], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0261, 0.0198, 0.0259, 0.0205, 0.0179, 0.0174, 0.0252], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:18:05,787 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22447.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:18:19,126 INFO [train.py:901] (1/4) Epoch 3, batch 6300, loss[loss=0.2906, simple_loss=0.3575, pruned_loss=0.1119, over 8101.00 frames. ], tot_loss[loss=0.3314, simple_loss=0.3796, pruned_loss=0.1416, over 1616811.12 frames. ], batch size: 23, lr: 2.17e-02, grad_scale: 16.0 +2023-02-05 22:18:36,440 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22492.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:18:39,306 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22496.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:18:54,108 INFO [train.py:901] (1/4) Epoch 3, batch 6350, loss[loss=0.3314, simple_loss=0.3786, pruned_loss=0.1421, over 8128.00 frames. ], tot_loss[loss=0.3287, simple_loss=0.3778, pruned_loss=0.1398, over 1618944.68 frames. ], batch size: 22, lr: 2.17e-02, grad_scale: 16.0 +2023-02-05 22:18:55,440 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.104e+02 3.537e+02 4.368e+02 5.315e+02 1.494e+03, threshold=8.736e+02, percent-clipped=5.0 +2023-02-05 22:18:57,018 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22521.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:08,839 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22538.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:28,745 INFO [train.py:901] (1/4) Epoch 3, batch 6400, loss[loss=0.3708, simple_loss=0.4164, pruned_loss=0.1626, over 8502.00 frames. ], tot_loss[loss=0.3267, simple_loss=0.3763, pruned_loss=0.1385, over 1617911.85 frames. ], batch size: 26, lr: 2.17e-02, grad_scale: 16.0 +2023-02-05 22:19:35,384 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22577.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:39,403 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8286, 2.3169, 2.0161, 2.9341, 1.4598, 1.3949, 1.8815, 2.3699], + device='cuda:1'), covar=tensor([0.1417, 0.1418, 0.1509, 0.0427, 0.1983, 0.2595, 0.1849, 0.1204], + device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0298, 0.0300, 0.0223, 0.0279, 0.0310, 0.0315, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 22:19:39,436 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22583.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:50,039 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22599.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:55,824 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22607.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:56,540 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22608.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:20:03,201 INFO [train.py:901] (1/4) Epoch 3, batch 6450, loss[loss=0.3456, simple_loss=0.3868, pruned_loss=0.1522, over 8017.00 frames. ], tot_loss[loss=0.328, simple_loss=0.3773, pruned_loss=0.1393, over 1619464.85 frames. ], batch size: 22, lr: 2.16e-02, grad_scale: 16.0 +2023-02-05 22:20:04,479 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 3.557e+02 4.436e+02 5.729e+02 1.082e+03, threshold=8.871e+02, percent-clipped=7.0 +2023-02-05 22:20:28,490 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22653.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:20:31,145 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22657.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:20:37,605 INFO [train.py:901] (1/4) Epoch 3, batch 6500, loss[loss=0.3865, simple_loss=0.4134, pruned_loss=0.1798, over 8492.00 frames. ], tot_loss[loss=0.3306, simple_loss=0.3792, pruned_loss=0.141, over 1618610.32 frames. ], batch size: 49, lr: 2.16e-02, grad_scale: 16.0 +2023-02-05 22:20:55,249 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22692.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:02,656 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22703.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:09,918 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22714.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:11,752 INFO [train.py:901] (1/4) Epoch 3, batch 6550, loss[loss=0.3779, simple_loss=0.4196, pruned_loss=0.1681, over 8558.00 frames. ], tot_loss[loss=0.3298, simple_loss=0.3789, pruned_loss=0.1404, over 1618925.66 frames. ], batch size: 34, lr: 2.16e-02, grad_scale: 16.0 +2023-02-05 22:21:13,166 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.155e+02 3.258e+02 3.883e+02 5.357e+02 1.264e+03, threshold=7.766e+02, percent-clipped=3.0 +2023-02-05 22:21:19,295 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22728.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:28,613 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-05 22:21:32,838 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22747.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:47,199 INFO [train.py:901] (1/4) Epoch 3, batch 6600, loss[loss=0.3769, simple_loss=0.4231, pruned_loss=0.1654, over 8600.00 frames. ], tot_loss[loss=0.3303, simple_loss=0.3789, pruned_loss=0.1409, over 1614008.26 frames. ], batch size: 39, lr: 2.16e-02, grad_scale: 8.0 +2023-02-05 22:21:47,899 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-05 22:21:54,955 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5094, 2.5932, 1.6707, 1.9506, 1.9835, 1.1416, 1.8068, 1.9400], + device='cuda:1'), covar=tensor([0.1147, 0.0326, 0.0840, 0.0724, 0.0677, 0.1198, 0.0926, 0.0806], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0231, 0.0322, 0.0315, 0.0328, 0.0314, 0.0340, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 22:22:19,032 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22812.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:22:22,348 INFO [train.py:901] (1/4) Epoch 3, batch 6650, loss[loss=0.3079, simple_loss=0.3734, pruned_loss=0.1213, over 8453.00 frames. ], tot_loss[loss=0.3301, simple_loss=0.3791, pruned_loss=0.1406, over 1620182.49 frames. ], batch size: 25, lr: 2.16e-02, grad_scale: 8.0 +2023-02-05 22:22:24,341 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.373e+02 3.456e+02 4.169e+02 5.335e+02 9.931e+02, threshold=8.339e+02, percent-clipped=8.0 +2023-02-05 22:22:40,061 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22843.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:22:53,741 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22862.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:22:54,497 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22863.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:22:57,072 INFO [train.py:901] (1/4) Epoch 3, batch 6700, loss[loss=0.3408, simple_loss=0.3799, pruned_loss=0.1508, over 7649.00 frames. ], tot_loss[loss=0.33, simple_loss=0.3786, pruned_loss=0.1407, over 1617554.46 frames. ], batch size: 19, lr: 2.15e-02, grad_scale: 8.0 +2023-02-05 22:23:12,655 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22888.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:23:26,895 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22909.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:23:32,716 INFO [train.py:901] (1/4) Epoch 3, batch 6750, loss[loss=0.3275, simple_loss=0.3679, pruned_loss=0.1435, over 7541.00 frames. ], tot_loss[loss=0.328, simple_loss=0.3774, pruned_loss=0.1393, over 1620418.41 frames. ], batch size: 18, lr: 2.15e-02, grad_scale: 8.0 +2023-02-05 22:23:34,753 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 3.597e+02 4.402e+02 5.483e+02 1.400e+03, threshold=8.804e+02, percent-clipped=7.0 +2023-02-05 22:23:44,511 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22934.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:23:53,620 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22948.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:05,894 INFO [train.py:901] (1/4) Epoch 3, batch 6800, loss[loss=0.348, simple_loss=0.3924, pruned_loss=0.1518, over 8444.00 frames. ], tot_loss[loss=0.3302, simple_loss=0.3788, pruned_loss=0.1408, over 1616427.61 frames. ], batch size: 27, lr: 2.15e-02, grad_scale: 8.0 +2023-02-05 22:24:05,909 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-05 22:24:08,782 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22970.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:10,818 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22973.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:26,137 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22995.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:30,239 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23001.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:41,423 INFO [train.py:901] (1/4) Epoch 3, batch 6850, loss[loss=0.3166, simple_loss=0.3745, pruned_loss=0.1294, over 8106.00 frames. ], tot_loss[loss=0.3291, simple_loss=0.3784, pruned_loss=0.1399, over 1616045.83 frames. ], batch size: 23, lr: 2.15e-02, grad_scale: 8.0 +2023-02-05 22:24:43,431 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 3.425e+02 4.505e+02 5.413e+02 1.323e+03, threshold=9.011e+02, percent-clipped=6.0 +2023-02-05 22:24:54,867 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-05 22:25:15,269 INFO [train.py:901] (1/4) Epoch 3, batch 6900, loss[loss=0.337, simple_loss=0.3838, pruned_loss=0.1451, over 8108.00 frames. ], tot_loss[loss=0.328, simple_loss=0.3773, pruned_loss=0.1393, over 1612777.54 frames. ], batch size: 23, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:25:50,193 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23116.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:25:50,685 INFO [train.py:901] (1/4) Epoch 3, batch 6950, loss[loss=0.3133, simple_loss=0.3654, pruned_loss=0.1306, over 8299.00 frames. ], tot_loss[loss=0.3274, simple_loss=0.3766, pruned_loss=0.1391, over 1613447.67 frames. ], batch size: 23, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:25:51,596 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23118.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:25:52,726 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 3.525e+02 4.440e+02 6.025e+02 1.140e+03, threshold=8.880e+02, percent-clipped=3.0 +2023-02-05 22:25:57,645 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23126.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:26:02,139 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-05 22:26:09,886 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23143.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:26:18,636 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23156.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:26:26,200 INFO [train.py:901] (1/4) Epoch 3, batch 7000, loss[loss=0.3309, simple_loss=0.3684, pruned_loss=0.1467, over 7643.00 frames. ], tot_loss[loss=0.3244, simple_loss=0.374, pruned_loss=0.1374, over 1610035.75 frames. ], batch size: 19, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:26:39,931 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23187.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:27:01,206 INFO [train.py:901] (1/4) Epoch 3, batch 7050, loss[loss=0.3268, simple_loss=0.379, pruned_loss=0.1373, over 8179.00 frames. ], tot_loss[loss=0.3266, simple_loss=0.376, pruned_loss=0.1386, over 1614822.89 frames. ], batch size: 23, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:27:03,863 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 3.682e+02 4.488e+02 5.424e+02 1.788e+03, threshold=8.977e+02, percent-clipped=6.0 +2023-02-05 22:27:14,103 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23235.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:27:22,748 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23247.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:27:36,329 INFO [train.py:901] (1/4) Epoch 3, batch 7100, loss[loss=0.3283, simple_loss=0.3822, pruned_loss=0.1372, over 7792.00 frames. ], tot_loss[loss=0.3246, simple_loss=0.3744, pruned_loss=0.1374, over 1612564.50 frames. ], batch size: 19, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:27:39,093 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23271.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:27:59,143 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23302.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:28:08,852 INFO [train.py:901] (1/4) Epoch 3, batch 7150, loss[loss=0.3145, simple_loss=0.3726, pruned_loss=0.1282, over 8573.00 frames. ], tot_loss[loss=0.3268, simple_loss=0.3762, pruned_loss=0.1387, over 1615571.80 frames. ], batch size: 34, lr: 2.13e-02, grad_scale: 8.0 +2023-02-05 22:28:09,679 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7597, 1.5442, 3.3514, 1.3705, 2.1871, 3.6968, 3.4340, 3.1967], + device='cuda:1'), covar=tensor([0.1186, 0.1403, 0.0369, 0.1821, 0.0765, 0.0233, 0.0323, 0.0530], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0264, 0.0206, 0.0263, 0.0204, 0.0181, 0.0182, 0.0253], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:28:10,872 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.197e+02 3.845e+02 4.572e+02 5.960e+02 1.048e+03, threshold=9.143e+02, percent-clipped=2.0 +2023-02-05 22:28:43,302 INFO [train.py:901] (1/4) Epoch 3, batch 7200, loss[loss=0.3757, simple_loss=0.4085, pruned_loss=0.1714, over 7274.00 frames. ], tot_loss[loss=0.3264, simple_loss=0.3761, pruned_loss=0.1384, over 1619093.10 frames. ], batch size: 71, lr: 2.13e-02, grad_scale: 8.0 +2023-02-05 22:28:47,006 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-05 22:28:47,558 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23372.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:29:04,755 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23397.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:29:17,772 INFO [train.py:901] (1/4) Epoch 3, batch 7250, loss[loss=0.3152, simple_loss=0.3678, pruned_loss=0.1313, over 7937.00 frames. ], tot_loss[loss=0.3278, simple_loss=0.3772, pruned_loss=0.1392, over 1619658.27 frames. ], batch size: 20, lr: 2.13e-02, grad_scale: 4.0 +2023-02-05 22:29:20,313 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.518e+02 3.505e+02 4.323e+02 5.847e+02 9.851e+02, threshold=8.646e+02, percent-clipped=2.0 +2023-02-05 22:29:52,896 INFO [train.py:901] (1/4) Epoch 3, batch 7300, loss[loss=0.2608, simple_loss=0.325, pruned_loss=0.09833, over 7545.00 frames. ], tot_loss[loss=0.3278, simple_loss=0.3773, pruned_loss=0.1391, over 1615466.75 frames. ], batch size: 18, lr: 2.13e-02, grad_scale: 4.0 +2023-02-05 22:29:55,065 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23470.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:30:02,087 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0474, 1.1774, 1.1756, 0.9188, 0.7921, 1.1703, 0.0355, 1.0604], + device='cuda:1'), covar=tensor([0.3084, 0.2082, 0.1388, 0.2052, 0.5612, 0.0978, 0.5176, 0.1530], + device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0116, 0.0085, 0.0159, 0.0186, 0.0080, 0.0151, 0.0117], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:30:21,339 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23506.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:30:28,671 INFO [train.py:901] (1/4) Epoch 3, batch 7350, loss[loss=0.3044, simple_loss=0.3608, pruned_loss=0.124, over 8104.00 frames. ], tot_loss[loss=0.3285, simple_loss=0.378, pruned_loss=0.1395, over 1618045.11 frames. ], batch size: 23, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:30:31,445 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.204e+02 3.295e+02 4.174e+02 5.897e+02 1.266e+03, threshold=8.348e+02, percent-clipped=6.0 +2023-02-05 22:30:35,728 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23527.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:30:45,653 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-05 22:30:52,307 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23552.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:30:56,349 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23558.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:31:03,075 INFO [train.py:901] (1/4) Epoch 3, batch 7400, loss[loss=0.3415, simple_loss=0.406, pruned_loss=0.1385, over 8196.00 frames. ], tot_loss[loss=0.3269, simple_loss=0.377, pruned_loss=0.1384, over 1617623.97 frames. ], batch size: 23, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:31:05,769 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-05 22:31:09,364 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5342, 2.2385, 3.8001, 1.1468, 2.5354, 1.8643, 1.6280, 2.0831], + device='cuda:1'), covar=tensor([0.1178, 0.1194, 0.0357, 0.2260, 0.1003, 0.1689, 0.1078, 0.1594], + device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0412, 0.0488, 0.0499, 0.0546, 0.0484, 0.0433, 0.0554], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 22:31:11,873 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23579.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:31:14,778 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23583.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:31:16,175 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:31:20,316 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23591.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:31:38,776 INFO [train.py:901] (1/4) Epoch 3, batch 7450, loss[loss=0.4341, simple_loss=0.443, pruned_loss=0.2126, over 7281.00 frames. ], tot_loss[loss=0.3257, simple_loss=0.3764, pruned_loss=0.1375, over 1616724.28 frames. ], batch size: 71, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:31:41,492 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.151e+02 3.560e+02 4.542e+02 5.434e+02 8.209e+02, threshold=9.083e+02, percent-clipped=0.0 +2023-02-05 22:31:44,189 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-05 22:32:11,862 INFO [train.py:901] (1/4) Epoch 3, batch 7500, loss[loss=0.3847, simple_loss=0.4271, pruned_loss=0.1711, over 8104.00 frames. ], tot_loss[loss=0.3282, simple_loss=0.3783, pruned_loss=0.139, over 1619522.52 frames. ], batch size: 23, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:32:31,249 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23694.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:32:39,259 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23706.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:32:47,011 INFO [train.py:901] (1/4) Epoch 3, batch 7550, loss[loss=0.3014, simple_loss=0.3633, pruned_loss=0.1198, over 8245.00 frames. ], tot_loss[loss=0.3278, simple_loss=0.3782, pruned_loss=0.1387, over 1623611.31 frames. ], batch size: 24, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:32:49,789 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.055e+02 3.573e+02 4.120e+02 5.568e+02 9.909e+02, threshold=8.240e+02, percent-clipped=1.0 +2023-02-05 22:33:21,007 INFO [train.py:901] (1/4) Epoch 3, batch 7600, loss[loss=0.2899, simple_loss=0.3368, pruned_loss=0.1215, over 7664.00 frames. ], tot_loss[loss=0.3262, simple_loss=0.3767, pruned_loss=0.1378, over 1619414.47 frames. ], batch size: 19, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:33:55,887 INFO [train.py:901] (1/4) Epoch 3, batch 7650, loss[loss=0.2607, simple_loss=0.3208, pruned_loss=0.1003, over 7694.00 frames. ], tot_loss[loss=0.3268, simple_loss=0.377, pruned_loss=0.1383, over 1613823.87 frames. ], batch size: 18, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:33:58,399 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.349e+02 3.333e+02 4.379e+02 5.791e+02 1.321e+03, threshold=8.759e+02, percent-clipped=7.0 +2023-02-05 22:34:07,042 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-05 22:34:12,885 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23841.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:34:15,491 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4061, 1.8914, 2.0128, 1.5297, 0.9814, 2.1475, 0.4151, 1.4371], + device='cuda:1'), covar=tensor([0.3642, 0.1730, 0.1263, 0.2689, 0.5679, 0.0872, 0.5579, 0.2186], + device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0116, 0.0083, 0.0159, 0.0188, 0.0080, 0.0147, 0.0117], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:34:19,387 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23850.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:34:30,199 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23866.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:34:30,652 INFO [train.py:901] (1/4) Epoch 3, batch 7700, loss[loss=0.2878, simple_loss=0.3582, pruned_loss=0.1087, over 8134.00 frames. ], tot_loss[loss=0.3252, simple_loss=0.3756, pruned_loss=0.1373, over 1612059.52 frames. ], batch size: 22, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:34:50,851 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-05 22:35:01,334 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-05 22:35:03,601 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8257, 3.7594, 3.4396, 1.7532, 3.3477, 3.3145, 3.5500, 3.0289], + device='cuda:1'), covar=tensor([0.0997, 0.0690, 0.1031, 0.4207, 0.0823, 0.0816, 0.1315, 0.0824], + device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0251, 0.0292, 0.0379, 0.0280, 0.0230, 0.0273, 0.0217], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 22:35:04,771 INFO [train.py:901] (1/4) Epoch 3, batch 7750, loss[loss=0.2619, simple_loss=0.3371, pruned_loss=0.09341, over 8129.00 frames. ], tot_loss[loss=0.3244, simple_loss=0.3751, pruned_loss=0.1368, over 1610661.35 frames. ], batch size: 22, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:35:08,099 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.077e+02 3.458e+02 4.167e+02 5.729e+02 1.393e+03, threshold=8.335e+02, percent-clipped=8.0 +2023-02-05 22:35:27,675 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23950.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:35:37,128 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23962.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:35:39,023 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23965.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:35:40,191 INFO [train.py:901] (1/4) Epoch 3, batch 7800, loss[loss=0.2983, simple_loss=0.3598, pruned_loss=0.1184, over 8028.00 frames. ], tot_loss[loss=0.3256, simple_loss=0.376, pruned_loss=0.1376, over 1619767.83 frames. ], batch size: 22, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:35:45,530 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23975.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:35:53,468 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23987.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:36:13,738 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 +2023-02-05 22:36:14,028 INFO [train.py:901] (1/4) Epoch 3, batch 7850, loss[loss=0.286, simple_loss=0.3523, pruned_loss=0.1098, over 8098.00 frames. ], tot_loss[loss=0.3268, simple_loss=0.3767, pruned_loss=0.1385, over 1619592.60 frames. ], batch size: 23, lr: 2.10e-02, grad_scale: 8.0 +2023-02-05 22:36:16,552 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.230e+02 3.608e+02 4.565e+02 5.801e+02 1.089e+03, threshold=9.129e+02, percent-clipped=5.0 +2023-02-05 22:36:39,254 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24055.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:36:47,360 INFO [train.py:901] (1/4) Epoch 3, batch 7900, loss[loss=0.3826, simple_loss=0.4098, pruned_loss=0.1777, over 8575.00 frames. ], tot_loss[loss=0.3251, simple_loss=0.3751, pruned_loss=0.1376, over 1617734.36 frames. ], batch size: 49, lr: 2.10e-02, grad_scale: 8.0 +2023-02-05 22:37:20,414 INFO [train.py:901] (1/4) Epoch 3, batch 7950, loss[loss=0.3505, simple_loss=0.3993, pruned_loss=0.1508, over 6912.00 frames. ], tot_loss[loss=0.3262, simple_loss=0.3757, pruned_loss=0.1383, over 1616461.90 frames. ], batch size: 71, lr: 2.10e-02, grad_scale: 8.0 +2023-02-05 22:37:23,172 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.080e+02 3.295e+02 4.369e+02 5.897e+02 2.335e+03, threshold=8.738e+02, percent-clipped=5.0 +2023-02-05 22:37:54,055 INFO [train.py:901] (1/4) Epoch 3, batch 8000, loss[loss=0.3996, simple_loss=0.4148, pruned_loss=0.1922, over 7360.00 frames. ], tot_loss[loss=0.3252, simple_loss=0.3749, pruned_loss=0.1377, over 1614689.10 frames. ], batch size: 74, lr: 2.10e-02, grad_scale: 8.0 +2023-02-05 22:38:27,963 INFO [train.py:901] (1/4) Epoch 3, batch 8050, loss[loss=0.2624, simple_loss=0.3141, pruned_loss=0.1053, over 7520.00 frames. ], tot_loss[loss=0.3244, simple_loss=0.373, pruned_loss=0.1379, over 1591144.32 frames. ], batch size: 18, lr: 2.09e-02, grad_scale: 8.0 +2023-02-05 22:38:30,754 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 3.328e+02 4.149e+02 5.404e+02 3.135e+03, threshold=8.298e+02, percent-clipped=6.0 +2023-02-05 22:38:30,988 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24221.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:38:48,120 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24246.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:39:03,851 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-05 22:39:07,719 INFO [train.py:901] (1/4) Epoch 4, batch 0, loss[loss=0.3027, simple_loss=0.3599, pruned_loss=0.1228, over 8124.00 frames. ], tot_loss[loss=0.3027, simple_loss=0.3599, pruned_loss=0.1228, over 8124.00 frames. ], batch size: 22, lr: 1.96e-02, grad_scale: 8.0 +2023-02-05 22:39:07,719 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 22:39:18,725 INFO [train.py:935] (1/4) Epoch 4, validation: loss=0.2476, simple_loss=0.3384, pruned_loss=0.07836, over 944034.00 frames. +2023-02-05 22:39:18,726 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-05 22:39:34,132 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-05 22:39:52,986 INFO [train.py:901] (1/4) Epoch 4, batch 50, loss[loss=0.3157, simple_loss=0.3536, pruned_loss=0.1389, over 7232.00 frames. ], tot_loss[loss=0.3242, simple_loss=0.3748, pruned_loss=0.1368, over 362988.95 frames. ], batch size: 16, lr: 1.96e-02, grad_scale: 8.0 +2023-02-05 22:40:07,592 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.017e+02 3.527e+02 4.250e+02 5.116e+02 9.987e+02, threshold=8.500e+02, percent-clipped=2.0 +2023-02-05 22:40:09,007 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-05 22:40:27,955 INFO [train.py:901] (1/4) Epoch 4, batch 100, loss[loss=0.3318, simple_loss=0.3874, pruned_loss=0.1381, over 8033.00 frames. ], tot_loss[loss=0.3213, simple_loss=0.373, pruned_loss=0.1348, over 642504.40 frames. ], batch size: 22, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:40:31,336 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-05 22:40:38,488 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-05 22:41:01,468 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=24399.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:41:01,649 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0997, 1.7980, 2.8776, 2.2962, 2.2914, 1.8523, 1.3423, 1.0236], + device='cuda:1'), covar=tensor([0.1164, 0.1285, 0.0253, 0.0524, 0.0521, 0.0658, 0.0801, 0.1247], + device='cuda:1'), in_proj_covar=tensor([0.0646, 0.0563, 0.0480, 0.0527, 0.0645, 0.0524, 0.0541, 0.0540], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:41:02,095 INFO [train.py:901] (1/4) Epoch 4, batch 150, loss[loss=0.3467, simple_loss=0.3879, pruned_loss=0.1527, over 7667.00 frames. ], tot_loss[loss=0.3182, simple_loss=0.3696, pruned_loss=0.1334, over 855774.98 frames. ], batch size: 19, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:41:14,385 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.68 vs. limit=5.0 +2023-02-05 22:41:17,166 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 3.490e+02 4.203e+02 5.614e+02 1.653e+03, threshold=8.406e+02, percent-clipped=4.0 +2023-02-05 22:41:37,212 INFO [train.py:901] (1/4) Epoch 4, batch 200, loss[loss=0.3005, simple_loss=0.3666, pruned_loss=0.1172, over 8603.00 frames. ], tot_loss[loss=0.3158, simple_loss=0.3686, pruned_loss=0.1315, over 1026505.81 frames. ], batch size: 31, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:42:11,045 INFO [train.py:901] (1/4) Epoch 4, batch 250, loss[loss=0.3759, simple_loss=0.4099, pruned_loss=0.1709, over 8431.00 frames. ], tot_loss[loss=0.3173, simple_loss=0.3697, pruned_loss=0.1325, over 1157764.87 frames. ], batch size: 27, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:42:20,367 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24514.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:42:23,564 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-05 22:42:24,844 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.156e+02 3.531e+02 4.434e+02 5.277e+02 1.190e+03, threshold=8.868e+02, percent-clipped=4.0 +2023-02-05 22:42:31,617 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-05 22:42:46,012 INFO [train.py:901] (1/4) Epoch 4, batch 300, loss[loss=0.2815, simple_loss=0.3397, pruned_loss=0.1117, over 7970.00 frames. ], tot_loss[loss=0.3199, simple_loss=0.3726, pruned_loss=0.1336, over 1265578.97 frames. ], batch size: 21, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:42:57,014 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24565.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:43:12,332 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24587.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:43:21,546 INFO [train.py:901] (1/4) Epoch 4, batch 350, loss[loss=0.292, simple_loss=0.3547, pruned_loss=0.1147, over 8252.00 frames. ], tot_loss[loss=0.3188, simple_loss=0.3715, pruned_loss=0.1331, over 1343364.53 frames. ], batch size: 22, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:43:35,592 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.777e+02 3.300e+02 4.421e+02 5.071e+02 1.044e+03, threshold=8.841e+02, percent-clipped=4.0 +2023-02-05 22:43:50,090 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-05 22:43:56,471 INFO [train.py:901] (1/4) Epoch 4, batch 400, loss[loss=0.3045, simple_loss=0.3511, pruned_loss=0.1289, over 8081.00 frames. ], tot_loss[loss=0.319, simple_loss=0.3719, pruned_loss=0.1331, over 1402379.10 frames. ], batch size: 21, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:44:12,739 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-05 22:44:15,945 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5763, 2.1244, 4.6559, 1.0685, 2.5972, 2.0816, 1.5495, 2.6546], + device='cuda:1'), covar=tensor([0.1878, 0.2100, 0.0655, 0.3569, 0.1487, 0.2303, 0.1835, 0.2127], + device='cuda:1'), in_proj_covar=tensor([0.0454, 0.0413, 0.0497, 0.0506, 0.0554, 0.0479, 0.0438, 0.0561], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 22:44:30,024 INFO [train.py:901] (1/4) Epoch 4, batch 450, loss[loss=0.2979, simple_loss=0.3617, pruned_loss=0.117, over 8585.00 frames. ], tot_loss[loss=0.3192, simple_loss=0.372, pruned_loss=0.1332, over 1452954.57 frames. ], batch size: 34, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:44:42,419 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-02-05 22:44:44,811 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.106e+02 3.414e+02 4.548e+02 5.600e+02 1.007e+03, threshold=9.096e+02, percent-clipped=5.0 +2023-02-05 22:45:04,972 INFO [train.py:901] (1/4) Epoch 4, batch 500, loss[loss=0.3051, simple_loss=0.3615, pruned_loss=0.1244, over 7816.00 frames. ], tot_loss[loss=0.3198, simple_loss=0.3726, pruned_loss=0.1335, over 1491801.73 frames. ], batch size: 20, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:45:19,904 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24770.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:45:28,214 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24783.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:45:36,833 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24795.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:45:40,046 INFO [train.py:901] (1/4) Epoch 4, batch 550, loss[loss=0.2686, simple_loss=0.3324, pruned_loss=0.1024, over 7540.00 frames. ], tot_loss[loss=0.3189, simple_loss=0.3721, pruned_loss=0.1329, over 1521523.53 frames. ], batch size: 18, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:45:53,858 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.817e+02 3.369e+02 4.426e+02 5.591e+02 8.767e+02, threshold=8.852e+02, percent-clipped=0.0 +2023-02-05 22:46:13,952 INFO [train.py:901] (1/4) Epoch 4, batch 600, loss[loss=0.3667, simple_loss=0.4041, pruned_loss=0.1646, over 8587.00 frames. ], tot_loss[loss=0.319, simple_loss=0.3723, pruned_loss=0.1329, over 1547734.94 frames. ], batch size: 31, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:46:24,940 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24866.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:46:28,941 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-05 22:46:39,228 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1528, 1.2782, 1.1739, 0.0904, 1.2044, 0.9847, 0.2066, 1.2656], + device='cuda:1'), covar=tensor([0.0093, 0.0072, 0.0063, 0.0155, 0.0096, 0.0253, 0.0203, 0.0079], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0175, 0.0140, 0.0219, 0.0165, 0.0297, 0.0236, 0.0207], + device='cuda:1'), out_proj_covar=tensor([1.0948e-04, 7.5440e-05, 5.9494e-05, 9.1786e-05, 7.2555e-05, 1.3915e-04, + 1.0301e-04, 8.7562e-05], device='cuda:1') +2023-02-05 22:46:49,154 INFO [train.py:901] (1/4) Epoch 4, batch 650, loss[loss=0.3354, simple_loss=0.3878, pruned_loss=0.1415, over 8581.00 frames. ], tot_loss[loss=0.3179, simple_loss=0.3713, pruned_loss=0.1322, over 1568945.33 frames. ], batch size: 31, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:46:55,185 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:47:00,774 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5191, 1.8428, 3.3586, 1.0018, 2.4169, 1.7103, 1.4743, 1.9451], + device='cuda:1'), covar=tensor([0.1824, 0.1884, 0.0646, 0.3371, 0.1309, 0.2505, 0.1827, 0.2230], + device='cuda:1'), in_proj_covar=tensor([0.0459, 0.0416, 0.0501, 0.0511, 0.0550, 0.0482, 0.0441, 0.0562], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 22:47:03,762 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.198e+02 3.310e+02 4.230e+02 5.108e+02 1.167e+03, threshold=8.459e+02, percent-clipped=4.0 +2023-02-05 22:47:10,606 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=24931.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:47:24,034 INFO [train.py:901] (1/4) Epoch 4, batch 700, loss[loss=0.3097, simple_loss=0.341, pruned_loss=0.1392, over 7809.00 frames. ], tot_loss[loss=0.318, simple_loss=0.3716, pruned_loss=0.1322, over 1584768.56 frames. ], batch size: 19, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:47:43,051 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1755, 3.7616, 3.3082, 4.2370, 2.1462, 2.6807, 2.2673, 3.8366], + device='cuda:1'), covar=tensor([0.0858, 0.1000, 0.1131, 0.0352, 0.1879, 0.1900, 0.2162, 0.0857], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0300, 0.0307, 0.0224, 0.0277, 0.0308, 0.0315, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:1') +2023-02-05 22:47:59,147 INFO [train.py:901] (1/4) Epoch 4, batch 750, loss[loss=0.313, simple_loss=0.3709, pruned_loss=0.1275, over 8392.00 frames. ], tot_loss[loss=0.3167, simple_loss=0.3708, pruned_loss=0.1313, over 1592351.71 frames. ], batch size: 49, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:48:08,255 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25013.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:48:13,341 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 3.175e+02 4.108e+02 5.247e+02 1.235e+03, threshold=8.217e+02, percent-clipped=4.0 +2023-02-05 22:48:14,034 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-05 22:48:15,510 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25024.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:48:22,610 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-05 22:48:30,677 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25046.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:48:33,224 INFO [train.py:901] (1/4) Epoch 4, batch 800, loss[loss=0.3158, simple_loss=0.3727, pruned_loss=0.1294, over 7980.00 frames. ], tot_loss[loss=0.3176, simple_loss=0.3709, pruned_loss=0.1322, over 1597176.77 frames. ], batch size: 21, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:49:06,964 INFO [train.py:901] (1/4) Epoch 4, batch 850, loss[loss=0.2245, simple_loss=0.295, pruned_loss=0.07702, over 7704.00 frames. ], tot_loss[loss=0.3178, simple_loss=0.3709, pruned_loss=0.1323, over 1599609.89 frames. ], batch size: 18, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:49:22,453 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 3.301e+02 4.277e+02 5.478e+02 1.022e+03, threshold=8.554e+02, percent-clipped=4.0 +2023-02-05 22:49:26,600 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25127.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:49:42,457 INFO [train.py:901] (1/4) Epoch 4, batch 900, loss[loss=0.2887, simple_loss=0.3588, pruned_loss=0.1093, over 8354.00 frames. ], tot_loss[loss=0.3162, simple_loss=0.3699, pruned_loss=0.1313, over 1604554.06 frames. ], batch size: 24, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:50:16,535 INFO [train.py:901] (1/4) Epoch 4, batch 950, loss[loss=0.323, simple_loss=0.3763, pruned_loss=0.1349, over 8463.00 frames. ], tot_loss[loss=0.3172, simple_loss=0.3698, pruned_loss=0.1323, over 1601822.89 frames. ], batch size: 25, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:50:20,252 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5011, 1.7640, 1.7051, 1.3669, 1.1382, 1.8844, 0.2709, 1.1645], + device='cuda:1'), covar=tensor([0.2680, 0.2383, 0.1803, 0.2712, 0.5630, 0.0917, 0.6368, 0.2651], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0115, 0.0085, 0.0159, 0.0189, 0.0082, 0.0154, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:50:23,515 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25210.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:50:30,874 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.215e+02 3.501e+02 4.488e+02 5.717e+02 1.063e+03, threshold=8.976e+02, percent-clipped=5.0 +2023-02-05 22:50:40,871 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-05 22:50:46,557 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25242.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:50:51,542 INFO [train.py:901] (1/4) Epoch 4, batch 1000, loss[loss=0.3749, simple_loss=0.4172, pruned_loss=0.1664, over 8327.00 frames. ], tot_loss[loss=0.3168, simple_loss=0.3692, pruned_loss=0.1322, over 1604810.20 frames. ], batch size: 26, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:51:12,232 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25280.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:51:13,313 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-05 22:51:25,827 INFO [train.py:901] (1/4) Epoch 4, batch 1050, loss[loss=0.3173, simple_loss=0.3589, pruned_loss=0.1379, over 8249.00 frames. ], tot_loss[loss=0.3169, simple_loss=0.3695, pruned_loss=0.1321, over 1606334.47 frames. ], batch size: 22, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:51:26,402 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-05 22:51:27,141 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25302.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:51:28,964 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25305.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:51:39,496 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 3.519e+02 4.399e+02 5.664e+02 1.146e+03, threshold=8.797e+02, percent-clipped=2.0 +2023-02-05 22:51:42,388 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25325.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:51:43,774 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25327.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:51:58,858 INFO [train.py:901] (1/4) Epoch 4, batch 1100, loss[loss=0.2684, simple_loss=0.3375, pruned_loss=0.0996, over 8457.00 frames. ], tot_loss[loss=0.3148, simple_loss=0.3678, pruned_loss=0.1309, over 1606499.14 frames. ], batch size: 25, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:52:04,505 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25357.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:52:34,558 INFO [train.py:901] (1/4) Epoch 4, batch 1150, loss[loss=0.3065, simple_loss=0.3783, pruned_loss=0.1173, over 8467.00 frames. ], tot_loss[loss=0.3151, simple_loss=0.3679, pruned_loss=0.1311, over 1606183.61 frames. ], batch size: 27, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:52:37,396 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-05 22:52:49,212 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.195e+02 3.278e+02 3.972e+02 4.649e+02 8.065e+02, threshold=7.944e+02, percent-clipped=0.0 +2023-02-05 22:52:59,737 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-02-05 22:53:06,097 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25446.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 22:53:08,506 INFO [train.py:901] (1/4) Epoch 4, batch 1200, loss[loss=0.3635, simple_loss=0.4095, pruned_loss=0.1588, over 8299.00 frames. ], tot_loss[loss=0.3155, simple_loss=0.3684, pruned_loss=0.1313, over 1606239.92 frames. ], batch size: 23, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:53:10,192 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-02-05 22:53:23,246 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25472.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:53:42,025 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25498.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:53:43,175 INFO [train.py:901] (1/4) Epoch 4, batch 1250, loss[loss=0.3319, simple_loss=0.3823, pruned_loss=0.1408, over 8328.00 frames. ], tot_loss[loss=0.3152, simple_loss=0.368, pruned_loss=0.1312, over 1608532.89 frames. ], batch size: 26, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:53:57,802 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 3.538e+02 4.328e+02 6.105e+02 1.271e+03, threshold=8.657e+02, percent-clipped=4.0 +2023-02-05 22:53:59,259 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25523.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:54:18,017 INFO [train.py:901] (1/4) Epoch 4, batch 1300, loss[loss=0.3452, simple_loss=0.3847, pruned_loss=0.1528, over 7796.00 frames. ], tot_loss[loss=0.315, simple_loss=0.3678, pruned_loss=0.1311, over 1608493.41 frames. ], batch size: 20, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:54:39,387 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25581.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:54:53,161 INFO [train.py:901] (1/4) Epoch 4, batch 1350, loss[loss=0.2735, simple_loss=0.3251, pruned_loss=0.111, over 7638.00 frames. ], tot_loss[loss=0.3145, simple_loss=0.3674, pruned_loss=0.1308, over 1606088.89 frames. ], batch size: 19, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:54:57,498 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25606.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:54:59,645 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3983, 1.5965, 1.6633, 1.3099, 0.8863, 1.6765, 0.1029, 0.9882], + device='cuda:1'), covar=tensor([0.2861, 0.1883, 0.1079, 0.2240, 0.6185, 0.0937, 0.5468, 0.2724], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0110, 0.0086, 0.0159, 0.0190, 0.0083, 0.0152, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 22:55:08,861 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 3.283e+02 4.098e+02 5.393e+02 1.175e+03, threshold=8.196e+02, percent-clipped=3.0 +2023-02-05 22:55:12,063 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.14 vs. limit=5.0 +2023-02-05 22:55:27,914 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=5.13 vs. limit=5.0 +2023-02-05 22:55:28,874 INFO [train.py:901] (1/4) Epoch 4, batch 1400, loss[loss=0.2824, simple_loss=0.3601, pruned_loss=0.1024, over 8449.00 frames. ], tot_loss[loss=0.3148, simple_loss=0.368, pruned_loss=0.1308, over 1608752.02 frames. ], batch size: 29, lr: 1.91e-02, grad_scale: 8.0 +2023-02-05 22:56:03,161 INFO [train.py:901] (1/4) Epoch 4, batch 1450, loss[loss=0.3396, simple_loss=0.3936, pruned_loss=0.1428, over 7962.00 frames. ], tot_loss[loss=0.3142, simple_loss=0.3681, pruned_loss=0.1301, over 1609724.95 frames. ], batch size: 21, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:56:05,848 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-05 22:56:18,905 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 3.243e+02 3.964e+02 4.847e+02 1.034e+03, threshold=7.929e+02, percent-clipped=2.0 +2023-02-05 22:56:23,200 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25728.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:56:38,604 INFO [train.py:901] (1/4) Epoch 4, batch 1500, loss[loss=0.3, simple_loss=0.3726, pruned_loss=0.1137, over 8497.00 frames. ], tot_loss[loss=0.3142, simple_loss=0.3686, pruned_loss=0.1299, over 1614866.24 frames. ], batch size: 26, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:56:40,774 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25753.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:57:02,625 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5395, 4.4821, 3.9490, 1.8611, 3.9654, 3.9659, 4.2954, 3.5885], + device='cuda:1'), covar=tensor([0.0879, 0.0504, 0.0816, 0.4455, 0.0722, 0.0656, 0.0980, 0.0715], + device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0254, 0.0300, 0.0381, 0.0294, 0.0236, 0.0280, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 22:57:05,990 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25790.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 22:57:12,582 INFO [train.py:901] (1/4) Epoch 4, batch 1550, loss[loss=0.2877, simple_loss=0.3613, pruned_loss=0.1071, over 8326.00 frames. ], tot_loss[loss=0.3134, simple_loss=0.3679, pruned_loss=0.1294, over 1616095.83 frames. ], batch size: 26, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:57:27,015 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.007e+02 3.100e+02 3.836e+02 5.066e+02 1.009e+03, threshold=7.672e+02, percent-clipped=5.0 +2023-02-05 22:57:31,370 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-02-05 22:57:46,741 INFO [train.py:901] (1/4) Epoch 4, batch 1600, loss[loss=0.2975, simple_loss=0.3501, pruned_loss=0.1224, over 7246.00 frames. ], tot_loss[loss=0.3134, simple_loss=0.3676, pruned_loss=0.1296, over 1616592.14 frames. ], batch size: 16, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:58:04,757 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25876.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 22:58:21,025 INFO [train.py:901] (1/4) Epoch 4, batch 1650, loss[loss=0.2932, simple_loss=0.3563, pruned_loss=0.1151, over 8191.00 frames. ], tot_loss[loss=0.3137, simple_loss=0.367, pruned_loss=0.1302, over 1606890.93 frames. ], batch size: 23, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:58:24,584 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25905.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 22:58:32,035 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4998, 3.1360, 2.4183, 4.0428, 1.7191, 1.6076, 2.1716, 3.5238], + device='cuda:1'), covar=tensor([0.0980, 0.1459, 0.1547, 0.0355, 0.1834, 0.2449, 0.2203, 0.0837], + device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0288, 0.0299, 0.0224, 0.0269, 0.0303, 0.0315, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-05 22:58:35,952 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.227e+02 3.823e+02 4.768e+02 5.766e+02 1.707e+03, threshold=9.535e+02, percent-clipped=9.0 +2023-02-05 22:58:56,114 INFO [train.py:901] (1/4) Epoch 4, batch 1700, loss[loss=0.2853, simple_loss=0.3494, pruned_loss=0.1106, over 8230.00 frames. ], tot_loss[loss=0.3134, simple_loss=0.3669, pruned_loss=0.1299, over 1612039.46 frames. ], batch size: 22, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:58:58,409 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4794, 2.0557, 1.3905, 2.6236, 1.4522, 1.0414, 1.5520, 2.0659], + device='cuda:1'), covar=tensor([0.1883, 0.1587, 0.3190, 0.0503, 0.1835, 0.3289, 0.1981, 0.1234], + device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0287, 0.0301, 0.0221, 0.0268, 0.0305, 0.0313, 0.0283], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-05 22:59:31,190 INFO [train.py:901] (1/4) Epoch 4, batch 1750, loss[loss=0.274, simple_loss=0.3339, pruned_loss=0.107, over 8655.00 frames. ], tot_loss[loss=0.3123, simple_loss=0.3664, pruned_loss=0.1291, over 1614823.59 frames. ], batch size: 34, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 22:59:47,004 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.961e+02 3.187e+02 3.816e+02 4.801e+02 8.317e+02, threshold=7.632e+02, percent-clipped=0.0 +2023-02-05 23:00:06,091 INFO [train.py:901] (1/4) Epoch 4, batch 1800, loss[loss=0.3121, simple_loss=0.3778, pruned_loss=0.1232, over 8245.00 frames. ], tot_loss[loss=0.313, simple_loss=0.3668, pruned_loss=0.1296, over 1616051.65 frames. ], batch size: 24, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 23:00:41,282 INFO [train.py:901] (1/4) Epoch 4, batch 1850, loss[loss=0.3292, simple_loss=0.3812, pruned_loss=0.1386, over 8542.00 frames. ], tot_loss[loss=0.3122, simple_loss=0.3665, pruned_loss=0.129, over 1617126.20 frames. ], batch size: 31, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 23:00:55,436 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26120.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:00:56,599 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.260e+02 3.379e+02 4.261e+02 5.084e+02 1.608e+03, threshold=8.521e+02, percent-clipped=6.0 +2023-02-05 23:01:15,415 INFO [train.py:901] (1/4) Epoch 4, batch 1900, loss[loss=0.2604, simple_loss=0.3143, pruned_loss=0.1033, over 7222.00 frames. ], tot_loss[loss=0.3124, simple_loss=0.3664, pruned_loss=0.1292, over 1615911.59 frames. ], batch size: 16, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 23:01:22,863 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26161.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:01:40,591 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26186.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:01:40,653 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26186.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:01:41,082 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-05 23:01:49,625 INFO [train.py:901] (1/4) Epoch 4, batch 1950, loss[loss=0.2867, simple_loss=0.3512, pruned_loss=0.1111, over 8097.00 frames. ], tot_loss[loss=0.3128, simple_loss=0.3671, pruned_loss=0.1293, over 1616843.06 frames. ], batch size: 23, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 23:01:52,424 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-05 23:02:04,046 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26220.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:02:05,152 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.303e+02 3.684e+02 4.572e+02 6.046e+02 1.247e+03, threshold=9.144e+02, percent-clipped=2.0 +2023-02-05 23:02:10,406 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-05 23:02:22,289 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-05 23:02:24,334 INFO [train.py:901] (1/4) Epoch 4, batch 2000, loss[loss=0.4108, simple_loss=0.4288, pruned_loss=0.1964, over 8251.00 frames. ], tot_loss[loss=0.3136, simple_loss=0.3674, pruned_loss=0.1299, over 1613666.21 frames. ], batch size: 24, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:02:36,624 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26268.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:02:59,487 INFO [train.py:901] (1/4) Epoch 4, batch 2050, loss[loss=0.298, simple_loss=0.3597, pruned_loss=0.1182, over 7957.00 frames. ], tot_loss[loss=0.313, simple_loss=0.3663, pruned_loss=0.1299, over 1609808.91 frames. ], batch size: 21, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:03:10,495 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4198, 2.0604, 1.9976, 1.1903, 2.0935, 1.3940, 0.6042, 1.7223], + device='cuda:1'), covar=tensor([0.0158, 0.0074, 0.0061, 0.0127, 0.0093, 0.0282, 0.0218, 0.0070], + device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0189, 0.0150, 0.0226, 0.0175, 0.0301, 0.0250, 0.0214], + device='cuda:1'), out_proj_covar=tensor([1.1004e-04, 7.9613e-05, 6.1679e-05, 9.1799e-05, 7.5091e-05, 1.3656e-04, + 1.0664e-04, 8.7591e-05], device='cuda:1') +2023-02-05 23:03:14,377 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.223e+02 3.433e+02 4.198e+02 5.260e+02 1.263e+03, threshold=8.396e+02, percent-clipped=5.0 +2023-02-05 23:03:24,276 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26335.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:03:34,740 INFO [train.py:901] (1/4) Epoch 4, batch 2100, loss[loss=0.3155, simple_loss=0.3724, pruned_loss=0.1293, over 8191.00 frames. ], tot_loss[loss=0.3142, simple_loss=0.3677, pruned_loss=0.1304, over 1613887.55 frames. ], batch size: 23, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:04:08,106 INFO [train.py:901] (1/4) Epoch 4, batch 2150, loss[loss=0.4336, simple_loss=0.4385, pruned_loss=0.2143, over 6637.00 frames. ], tot_loss[loss=0.3133, simple_loss=0.3673, pruned_loss=0.1297, over 1616944.47 frames. ], batch size: 71, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:04:13,091 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5789, 1.8341, 3.2589, 1.1409, 2.2109, 1.9042, 1.4656, 1.9182], + device='cuda:1'), covar=tensor([0.1263, 0.1448, 0.0444, 0.2663, 0.1126, 0.1927, 0.1414, 0.1818], + device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0415, 0.0505, 0.0504, 0.0546, 0.0487, 0.0439, 0.0554], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:04:24,387 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.431e+02 3.407e+02 4.210e+02 5.616e+02 1.521e+03, threshold=8.419e+02, percent-clipped=4.0 +2023-02-05 23:04:31,148 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26432.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:04:43,606 INFO [train.py:901] (1/4) Epoch 4, batch 2200, loss[loss=0.3355, simple_loss=0.3719, pruned_loss=0.1495, over 8134.00 frames. ], tot_loss[loss=0.3118, simple_loss=0.3661, pruned_loss=0.1287, over 1620580.60 frames. ], batch size: 22, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:04:53,163 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26464.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:04:55,297 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1892, 1.9697, 1.7329, 1.9174, 1.8576, 1.7403, 2.7881, 2.0291], + device='cuda:1'), covar=tensor([0.0600, 0.1294, 0.1861, 0.1334, 0.0739, 0.1596, 0.0734, 0.0594], + device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0199, 0.0240, 0.0201, 0.0162, 0.0206, 0.0168, 0.0167], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0006, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:1') +2023-02-05 23:05:18,122 INFO [train.py:901] (1/4) Epoch 4, batch 2250, loss[loss=0.3041, simple_loss=0.3584, pruned_loss=0.1249, over 7772.00 frames. ], tot_loss[loss=0.3101, simple_loss=0.3652, pruned_loss=0.1275, over 1621273.35 frames. ], batch size: 19, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:05:32,594 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4879, 1.3490, 2.7612, 1.1156, 2.0719, 2.9640, 2.9243, 2.5343], + device='cuda:1'), covar=tensor([0.0994, 0.1297, 0.0404, 0.1961, 0.0596, 0.0356, 0.0427, 0.0675], + device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0257, 0.0207, 0.0262, 0.0207, 0.0188, 0.0196, 0.0259], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:05:33,086 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.934e+02 3.188e+02 3.857e+02 4.748e+02 9.287e+02, threshold=7.714e+02, percent-clipped=1.0 +2023-02-05 23:05:38,940 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26530.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:05:52,824 INFO [train.py:901] (1/4) Epoch 4, batch 2300, loss[loss=0.3125, simple_loss=0.372, pruned_loss=0.1265, over 8493.00 frames. ], tot_loss[loss=0.3099, simple_loss=0.3645, pruned_loss=0.1277, over 1613403.64 frames. ], batch size: 26, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:06:02,938 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3841, 2.0704, 3.1176, 2.7400, 2.5191, 1.9822, 1.3159, 1.2584], + device='cuda:1'), covar=tensor([0.1349, 0.1545, 0.0291, 0.0597, 0.0757, 0.0699, 0.0929, 0.1536], + device='cuda:1'), in_proj_covar=tensor([0.0667, 0.0595, 0.0499, 0.0566, 0.0687, 0.0546, 0.0549, 0.0563], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 23:06:12,941 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26579.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:06:21,882 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26591.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:06:27,543 INFO [train.py:901] (1/4) Epoch 4, batch 2350, loss[loss=0.299, simple_loss=0.3672, pruned_loss=0.1153, over 8190.00 frames. ], tot_loss[loss=0.3095, simple_loss=0.3641, pruned_loss=0.1274, over 1610252.15 frames. ], batch size: 23, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:06:35,939 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26612.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:06:38,734 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26616.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:06:42,411 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.224e+02 3.505e+02 4.841e+02 5.770e+02 1.247e+03, threshold=9.683e+02, percent-clipped=6.0 +2023-02-05 23:06:58,650 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26645.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:07:01,512 INFO [train.py:901] (1/4) Epoch 4, batch 2400, loss[loss=0.2733, simple_loss=0.3245, pruned_loss=0.1111, over 7405.00 frames. ], tot_loss[loss=0.3114, simple_loss=0.3655, pruned_loss=0.1286, over 1612281.64 frames. ], batch size: 17, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:07:03,752 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7816, 1.3288, 5.9180, 2.2683, 5.2664, 4.8382, 5.4161, 5.4159], + device='cuda:1'), covar=tensor([0.0478, 0.3964, 0.0243, 0.1981, 0.0733, 0.0459, 0.0390, 0.0374], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0442, 0.0342, 0.0357, 0.0419, 0.0359, 0.0341, 0.0378], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-05 23:07:37,131 INFO [train.py:901] (1/4) Epoch 4, batch 2450, loss[loss=0.3062, simple_loss=0.3654, pruned_loss=0.1235, over 8041.00 frames. ], tot_loss[loss=0.3128, simple_loss=0.3662, pruned_loss=0.1297, over 1610296.90 frames. ], batch size: 22, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:07:47,372 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1525, 1.8385, 3.4249, 0.8742, 2.1868, 1.4580, 1.3249, 1.8828], + device='cuda:1'), covar=tensor([0.2280, 0.2200, 0.0815, 0.4104, 0.1709, 0.2975, 0.2136, 0.2535], + device='cuda:1'), in_proj_covar=tensor([0.0451, 0.0420, 0.0506, 0.0511, 0.0554, 0.0483, 0.0439, 0.0561], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:07:51,851 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 3.211e+02 4.300e+02 5.616e+02 1.854e+03, threshold=8.599e+02, percent-clipped=7.0 +2023-02-05 23:07:55,327 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26727.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:07:58,057 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3962, 2.0766, 1.9923, 0.7973, 2.1005, 1.5361, 0.5870, 1.7961], + device='cuda:1'), covar=tensor([0.0168, 0.0072, 0.0078, 0.0148, 0.0102, 0.0257, 0.0239, 0.0072], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0188, 0.0154, 0.0226, 0.0172, 0.0307, 0.0251, 0.0210], + device='cuda:1'), out_proj_covar=tensor([1.1017e-04, 7.8155e-05, 6.2722e-05, 9.1577e-05, 7.2669e-05, 1.3801e-04, + 1.0628e-04, 8.5724e-05], device='cuda:1') +2023-02-05 23:08:10,589 INFO [train.py:901] (1/4) Epoch 4, batch 2500, loss[loss=0.3214, simple_loss=0.3807, pruned_loss=0.131, over 8360.00 frames. ], tot_loss[loss=0.3124, simple_loss=0.3659, pruned_loss=0.1295, over 1604712.06 frames. ], batch size: 24, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:08:19,999 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6413, 1.4734, 3.1291, 1.2792, 2.1361, 3.4200, 3.2503, 2.8166], + device='cuda:1'), covar=tensor([0.1038, 0.1363, 0.0311, 0.1919, 0.0720, 0.0232, 0.0342, 0.0608], + device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0259, 0.0205, 0.0260, 0.0209, 0.0186, 0.0192, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:08:26,662 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8132, 1.6252, 3.4988, 1.3866, 2.4282, 3.9200, 3.7122, 3.2998], + device='cuda:1'), covar=tensor([0.1108, 0.1458, 0.0271, 0.1900, 0.0702, 0.0227, 0.0308, 0.0591], + device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0261, 0.0207, 0.0262, 0.0211, 0.0187, 0.0193, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:08:29,253 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26776.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:08:45,144 INFO [train.py:901] (1/4) Epoch 4, batch 2550, loss[loss=0.3197, simple_loss=0.3832, pruned_loss=0.1281, over 8451.00 frames. ], tot_loss[loss=0.3116, simple_loss=0.3654, pruned_loss=0.1289, over 1603138.38 frames. ], batch size: 27, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:09:01,297 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.235e+02 3.301e+02 4.146e+02 5.074e+02 1.055e+03, threshold=8.293e+02, percent-clipped=2.0 +2023-02-05 23:09:05,071 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1014, 1.6767, 2.9394, 2.4081, 2.3171, 1.7503, 1.2831, 1.0179], + device='cuda:1'), covar=tensor([0.1436, 0.1787, 0.0282, 0.0623, 0.0660, 0.0991, 0.1074, 0.1442], + device='cuda:1'), in_proj_covar=tensor([0.0661, 0.0593, 0.0492, 0.0567, 0.0677, 0.0549, 0.0546, 0.0557], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 23:09:10,584 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26835.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:09:20,517 INFO [train.py:901] (1/4) Epoch 4, batch 2600, loss[loss=0.3042, simple_loss=0.3609, pruned_loss=0.1238, over 7807.00 frames. ], tot_loss[loss=0.312, simple_loss=0.3663, pruned_loss=0.1289, over 1608021.71 frames. ], batch size: 20, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:09:27,745 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26860.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:09:50,362 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26891.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:09:56,432 INFO [train.py:901] (1/4) Epoch 4, batch 2650, loss[loss=0.3081, simple_loss=0.3744, pruned_loss=0.121, over 8323.00 frames. ], tot_loss[loss=0.3098, simple_loss=0.3648, pruned_loss=0.1274, over 1610301.83 frames. ], batch size: 25, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:09:57,281 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26901.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:10:12,338 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 3.245e+02 3.916e+02 5.024e+02 1.006e+03, threshold=7.831e+02, percent-clipped=3.0 +2023-02-05 23:10:12,505 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26922.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:10:15,324 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26926.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:10:32,098 INFO [train.py:901] (1/4) Epoch 4, batch 2700, loss[loss=0.2977, simple_loss=0.3724, pruned_loss=0.1115, over 8612.00 frames. ], tot_loss[loss=0.3089, simple_loss=0.3646, pruned_loss=0.1265, over 1615480.95 frames. ], batch size: 31, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:10:44,358 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26968.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:10:54,318 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26983.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:11:00,901 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4570, 4.5550, 4.0749, 1.7308, 3.9729, 3.9193, 4.2015, 3.4376], + device='cuda:1'), covar=tensor([0.0805, 0.0518, 0.0973, 0.4356, 0.0648, 0.0649, 0.1177, 0.0729], + device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0252, 0.0302, 0.0383, 0.0288, 0.0233, 0.0282, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 23:11:05,901 INFO [train.py:901] (1/4) Epoch 4, batch 2750, loss[loss=0.361, simple_loss=0.4074, pruned_loss=0.1573, over 8479.00 frames. ], tot_loss[loss=0.3111, simple_loss=0.3659, pruned_loss=0.1282, over 1612424.13 frames. ], batch size: 29, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:11:10,046 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-05 23:11:12,356 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27008.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:11:21,386 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.241e+02 3.589e+02 4.354e+02 5.460e+02 1.197e+03, threshold=8.707e+02, percent-clipped=9.0 +2023-02-05 23:11:40,850 INFO [train.py:901] (1/4) Epoch 4, batch 2800, loss[loss=0.307, simple_loss=0.3772, pruned_loss=0.1184, over 8461.00 frames. ], tot_loss[loss=0.3118, simple_loss=0.3664, pruned_loss=0.1286, over 1614383.95 frames. ], batch size: 25, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:11:41,014 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3708, 1.4252, 3.0999, 1.3267, 2.1737, 3.3369, 3.0970, 2.8439], + device='cuda:1'), covar=tensor([0.1133, 0.1384, 0.0366, 0.1949, 0.0699, 0.0301, 0.0430, 0.0614], + device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0257, 0.0210, 0.0262, 0.0211, 0.0187, 0.0195, 0.0262], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:11:52,405 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5673, 1.7441, 1.8427, 1.5573, 0.8647, 1.8084, 0.2392, 1.1714], + device='cuda:1'), covar=tensor([0.3814, 0.2764, 0.1178, 0.2075, 0.6938, 0.1213, 0.7083, 0.2494], + device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0114, 0.0083, 0.0162, 0.0200, 0.0084, 0.0155, 0.0121], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 23:12:09,983 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-02-05 23:12:14,847 INFO [train.py:901] (1/4) Epoch 4, batch 2850, loss[loss=0.2575, simple_loss=0.3234, pruned_loss=0.09579, over 7969.00 frames. ], tot_loss[loss=0.3121, simple_loss=0.3667, pruned_loss=0.1288, over 1620512.28 frames. ], batch size: 21, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:12:30,251 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 3.374e+02 4.464e+02 5.831e+02 1.992e+03, threshold=8.927e+02, percent-clipped=6.0 +2023-02-05 23:12:47,565 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27147.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:12:49,278 INFO [train.py:901] (1/4) Epoch 4, batch 2900, loss[loss=0.3115, simple_loss=0.3745, pruned_loss=0.1243, over 8664.00 frames. ], tot_loss[loss=0.3112, simple_loss=0.3657, pruned_loss=0.1284, over 1613012.86 frames. ], batch size: 34, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:13:00,618 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27166.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:13:05,274 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27172.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:13:11,781 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-05 23:13:24,074 INFO [train.py:901] (1/4) Epoch 4, batch 2950, loss[loss=0.3614, simple_loss=0.3798, pruned_loss=0.1715, over 6360.00 frames. ], tot_loss[loss=0.3132, simple_loss=0.3666, pruned_loss=0.1299, over 1607920.41 frames. ], batch size: 14, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:13:24,412 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6191, 2.0445, 2.0802, 0.8749, 2.1639, 1.4747, 0.5836, 1.7346], + device='cuda:1'), covar=tensor([0.0194, 0.0104, 0.0087, 0.0187, 0.0150, 0.0277, 0.0292, 0.0090], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0195, 0.0158, 0.0230, 0.0187, 0.0314, 0.0257, 0.0214], + device='cuda:1'), out_proj_covar=tensor([1.1159e-04, 8.0623e-05, 6.3440e-05, 9.2207e-05, 7.8582e-05, 1.3990e-04, + 1.0875e-04, 8.6472e-05], device='cuda:1') +2023-02-05 23:13:38,737 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.890e+02 3.092e+02 3.649e+02 5.055e+02 1.216e+03, threshold=7.299e+02, percent-clipped=3.0 +2023-02-05 23:13:58,839 INFO [train.py:901] (1/4) Epoch 4, batch 3000, loss[loss=0.3256, simple_loss=0.3787, pruned_loss=0.1362, over 8621.00 frames. ], tot_loss[loss=0.3108, simple_loss=0.3645, pruned_loss=0.1286, over 1607196.03 frames. ], batch size: 31, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:13:58,839 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 23:14:11,268 INFO [train.py:935] (1/4) Epoch 4, validation: loss=0.2374, simple_loss=0.3304, pruned_loss=0.07225, over 944034.00 frames. +2023-02-05 23:14:11,269 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-05 23:14:23,026 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27266.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:14:45,404 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-05 23:14:45,710 INFO [train.py:901] (1/4) Epoch 4, batch 3050, loss[loss=0.3212, simple_loss=0.3607, pruned_loss=0.1409, over 8528.00 frames. ], tot_loss[loss=0.3109, simple_loss=0.3642, pruned_loss=0.1289, over 1603825.96 frames. ], batch size: 28, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:14:54,660 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27312.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:15:01,936 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 3.415e+02 4.317e+02 5.768e+02 1.933e+03, threshold=8.634e+02, percent-clipped=10.0 +2023-02-05 23:15:20,627 INFO [train.py:901] (1/4) Epoch 4, batch 3100, loss[loss=0.2922, simple_loss=0.3482, pruned_loss=0.1181, over 7802.00 frames. ], tot_loss[loss=0.3103, simple_loss=0.3639, pruned_loss=0.1283, over 1604910.81 frames. ], batch size: 20, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:15:41,821 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27381.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:15:54,812 INFO [train.py:901] (1/4) Epoch 4, batch 3150, loss[loss=0.3022, simple_loss=0.3556, pruned_loss=0.1244, over 7978.00 frames. ], tot_loss[loss=0.3095, simple_loss=0.3637, pruned_loss=0.1276, over 1605526.69 frames. ], batch size: 21, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:16:09,479 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 3.237e+02 4.041e+02 5.193e+02 1.210e+03, threshold=8.082e+02, percent-clipped=3.0 +2023-02-05 23:16:13,662 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27427.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:16:16,326 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2941, 1.7048, 2.9516, 1.0172, 1.8908, 1.7134, 1.3822, 1.5440], + device='cuda:1'), covar=tensor([0.1328, 0.1361, 0.0485, 0.2571, 0.1153, 0.1935, 0.1188, 0.1741], + device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0417, 0.0499, 0.0512, 0.0560, 0.0494, 0.0430, 0.0560], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:16:29,604 INFO [train.py:901] (1/4) Epoch 4, batch 3200, loss[loss=0.3166, simple_loss=0.3711, pruned_loss=0.131, over 8090.00 frames. ], tot_loss[loss=0.3103, simple_loss=0.3648, pruned_loss=0.1279, over 1609137.61 frames. ], batch size: 21, lr: 1.84e-02, grad_scale: 8.0 +2023-02-05 23:16:33,644 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0873, 4.1067, 3.6518, 1.8316, 3.6605, 3.4899, 3.7182, 3.0053], + device='cuda:1'), covar=tensor([0.0856, 0.0520, 0.0894, 0.4258, 0.0724, 0.0774, 0.1113, 0.0871], + device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0259, 0.0300, 0.0383, 0.0289, 0.0238, 0.0281, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 23:16:50,826 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1031, 3.0274, 2.7451, 1.6183, 2.7129, 2.6642, 2.8442, 2.4452], + device='cuda:1'), covar=tensor([0.1325, 0.0926, 0.1195, 0.4513, 0.1106, 0.1049, 0.1490, 0.1059], + device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0260, 0.0301, 0.0383, 0.0290, 0.0238, 0.0282, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 23:17:03,110 INFO [train.py:901] (1/4) Epoch 4, batch 3250, loss[loss=0.2973, simple_loss=0.3484, pruned_loss=0.1231, over 6806.00 frames. ], tot_loss[loss=0.309, simple_loss=0.3634, pruned_loss=0.1273, over 1609370.68 frames. ], batch size: 15, lr: 1.84e-02, grad_scale: 8.0 +2023-02-05 23:17:07,516 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-05 23:17:10,578 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27510.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:17:18,418 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 3.449e+02 4.059e+02 4.930e+02 7.939e+02, threshold=8.117e+02, percent-clipped=0.0 +2023-02-05 23:17:37,474 INFO [train.py:901] (1/4) Epoch 4, batch 3300, loss[loss=0.3104, simple_loss=0.3341, pruned_loss=0.1434, over 7546.00 frames. ], tot_loss[loss=0.3092, simple_loss=0.3641, pruned_loss=0.1271, over 1612454.93 frames. ], batch size: 18, lr: 1.84e-02, grad_scale: 8.0 +2023-02-05 23:18:12,284 INFO [train.py:901] (1/4) Epoch 4, batch 3350, loss[loss=0.2815, simple_loss=0.3486, pruned_loss=0.1072, over 8139.00 frames. ], tot_loss[loss=0.3094, simple_loss=0.3646, pruned_loss=0.1271, over 1617788.18 frames. ], batch size: 22, lr: 1.84e-02, grad_scale: 8.0 +2023-02-05 23:18:28,394 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.390e+02 3.326e+02 4.176e+02 5.439e+02 1.733e+03, threshold=8.353e+02, percent-clipped=9.0 +2023-02-05 23:18:30,489 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27625.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:18:38,365 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27637.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:18:46,891 INFO [train.py:901] (1/4) Epoch 4, batch 3400, loss[loss=0.3077, simple_loss=0.3481, pruned_loss=0.1336, over 7508.00 frames. ], tot_loss[loss=0.3121, simple_loss=0.366, pruned_loss=0.1291, over 1614411.47 frames. ], batch size: 18, lr: 1.84e-02, grad_scale: 16.0 +2023-02-05 23:18:55,791 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27662.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:19:10,319 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27683.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:19:21,476 INFO [train.py:901] (1/4) Epoch 4, batch 3450, loss[loss=0.2616, simple_loss=0.3267, pruned_loss=0.09823, over 7812.00 frames. ], tot_loss[loss=0.3116, simple_loss=0.366, pruned_loss=0.1287, over 1613799.15 frames. ], batch size: 20, lr: 1.84e-02, grad_scale: 16.0 +2023-02-05 23:19:26,943 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27708.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:19:36,065 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.499e+02 3.357e+02 4.072e+02 5.275e+02 9.264e+02, threshold=8.144e+02, percent-clipped=1.0 +2023-02-05 23:19:43,548 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27732.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:19:46,954 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.85 vs. limit=5.0 +2023-02-05 23:19:55,911 INFO [train.py:901] (1/4) Epoch 4, batch 3500, loss[loss=0.3375, simple_loss=0.389, pruned_loss=0.143, over 8617.00 frames. ], tot_loss[loss=0.3126, simple_loss=0.3669, pruned_loss=0.1291, over 1616457.62 frames. ], batch size: 34, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:20:10,690 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-05 23:20:21,912 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-05 23:20:31,120 INFO [train.py:901] (1/4) Epoch 4, batch 3550, loss[loss=0.4197, simple_loss=0.4381, pruned_loss=0.2006, over 6560.00 frames. ], tot_loss[loss=0.3111, simple_loss=0.3657, pruned_loss=0.1283, over 1612593.48 frames. ], batch size: 71, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:20:46,083 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.197e+02 3.262e+02 3.955e+02 5.254e+02 1.114e+03, threshold=7.909e+02, percent-clipped=8.0 +2023-02-05 23:21:05,471 INFO [train.py:901] (1/4) Epoch 4, batch 3600, loss[loss=0.3176, simple_loss=0.3601, pruned_loss=0.1376, over 7775.00 frames. ], tot_loss[loss=0.312, simple_loss=0.366, pruned_loss=0.129, over 1612530.32 frames. ], batch size: 19, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:21:27,350 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27881.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:21:39,936 INFO [train.py:901] (1/4) Epoch 4, batch 3650, loss[loss=0.2902, simple_loss=0.3427, pruned_loss=0.1188, over 8026.00 frames. ], tot_loss[loss=0.3115, simple_loss=0.3659, pruned_loss=0.1286, over 1611909.16 frames. ], batch size: 22, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:21:44,917 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27906.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:21:56,104 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.998e+02 3.334e+02 3.945e+02 4.811e+02 1.062e+03, threshold=7.891e+02, percent-clipped=4.0 +2023-02-05 23:22:13,494 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-05 23:22:14,794 INFO [train.py:901] (1/4) Epoch 4, batch 3700, loss[loss=0.3366, simple_loss=0.3762, pruned_loss=0.1485, over 7812.00 frames. ], tot_loss[loss=0.312, simple_loss=0.366, pruned_loss=0.129, over 1614678.61 frames. ], batch size: 20, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:22:42,571 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.3679, 1.9889, 1.5665, 1.4371, 1.7155, 1.7495, 2.2566, 2.2669], + device='cuda:1'), covar=tensor([0.0501, 0.1255, 0.1826, 0.1430, 0.0714, 0.1512, 0.0749, 0.0504], + device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0196, 0.0236, 0.0197, 0.0154, 0.0200, 0.0160, 0.0165], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0007, 0.0007, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:1') +2023-02-05 23:22:49,607 INFO [train.py:901] (1/4) Epoch 4, batch 3750, loss[loss=0.3115, simple_loss=0.3711, pruned_loss=0.1259, over 8082.00 frames. ], tot_loss[loss=0.3114, simple_loss=0.3655, pruned_loss=0.1287, over 1610583.52 frames. ], batch size: 21, lr: 1.83e-02, grad_scale: 8.0 +2023-02-05 23:23:05,800 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.245e+02 3.553e+02 4.442e+02 6.055e+02 1.985e+03, threshold=8.883e+02, percent-clipped=11.0 +2023-02-05 23:23:20,799 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.17 vs. limit=2.0 +2023-02-05 23:23:25,245 INFO [train.py:901] (1/4) Epoch 4, batch 3800, loss[loss=0.3666, simple_loss=0.4121, pruned_loss=0.1605, over 8452.00 frames. ], tot_loss[loss=0.3114, simple_loss=0.3651, pruned_loss=0.1289, over 1605269.99 frames. ], batch size: 27, lr: 1.83e-02, grad_scale: 8.0 +2023-02-05 23:23:33,493 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3308, 2.1294, 1.5853, 2.0775, 1.7920, 1.3143, 1.4412, 1.8961], + device='cuda:1'), covar=tensor([0.0988, 0.0391, 0.0908, 0.0406, 0.0630, 0.1101, 0.0886, 0.0626], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0246, 0.0309, 0.0304, 0.0332, 0.0319, 0.0341, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 23:23:38,933 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3190, 1.9002, 3.1073, 2.3560, 2.5612, 1.9587, 1.3769, 1.2094], + device='cuda:1'), covar=tensor([0.1432, 0.1573, 0.0339, 0.0741, 0.0666, 0.0788, 0.0913, 0.1611], + device='cuda:1'), in_proj_covar=tensor([0.0683, 0.0610, 0.0519, 0.0573, 0.0690, 0.0564, 0.0569, 0.0575], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 23:23:42,794 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28076.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:23:45,521 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8998, 1.5045, 3.3752, 1.3546, 2.2829, 3.8277, 3.5844, 3.2419], + device='cuda:1'), covar=tensor([0.1199, 0.1573, 0.0387, 0.2072, 0.0791, 0.0258, 0.0471, 0.0743], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0266, 0.0218, 0.0264, 0.0218, 0.0194, 0.0201, 0.0274], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:23:47,597 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7622, 1.9304, 2.1118, 0.8413, 2.1514, 1.5209, 0.6127, 1.8523], + device='cuda:1'), covar=tensor([0.0144, 0.0088, 0.0080, 0.0173, 0.0134, 0.0281, 0.0262, 0.0080], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0197, 0.0155, 0.0229, 0.0185, 0.0315, 0.0253, 0.0218], + device='cuda:1'), out_proj_covar=tensor([1.0771e-04, 8.0089e-05, 6.0441e-05, 9.0932e-05, 7.6607e-05, 1.3796e-04, + 1.0568e-04, 8.6719e-05], device='cuda:1') +2023-02-05 23:24:00,306 INFO [train.py:901] (1/4) Epoch 4, batch 3850, loss[loss=0.3027, simple_loss=0.3567, pruned_loss=0.1243, over 8109.00 frames. ], tot_loss[loss=0.3119, simple_loss=0.3655, pruned_loss=0.1291, over 1607755.24 frames. ], batch size: 23, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:24:14,107 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2670, 1.5823, 1.5491, 1.2597, 0.8440, 1.6859, 0.0602, 0.9382], + device='cuda:1'), covar=tensor([0.3544, 0.1783, 0.1595, 0.2456, 0.6155, 0.0856, 0.6003, 0.2525], + device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0110, 0.0083, 0.0163, 0.0193, 0.0080, 0.0147, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 23:24:15,225 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.006e+02 3.238e+02 4.124e+02 5.182e+02 9.210e+02, threshold=8.247e+02, percent-clipped=1.0 +2023-02-05 23:24:17,313 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-05 23:24:27,548 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5534, 2.7546, 1.6782, 2.1378, 2.4069, 1.3712, 1.6463, 2.0548], + device='cuda:1'), covar=tensor([0.1205, 0.0253, 0.0802, 0.0568, 0.0475, 0.1156, 0.1023, 0.0806], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0239, 0.0306, 0.0306, 0.0326, 0.0319, 0.0342, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 23:24:32,277 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4485, 2.0221, 3.3282, 0.9892, 2.2334, 1.6186, 1.4886, 1.8832], + device='cuda:1'), covar=tensor([0.1477, 0.1535, 0.0607, 0.3039, 0.1415, 0.2389, 0.1364, 0.2192], + device='cuda:1'), in_proj_covar=tensor([0.0453, 0.0427, 0.0503, 0.0513, 0.0563, 0.0493, 0.0435, 0.0564], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:24:34,698 INFO [train.py:901] (1/4) Epoch 4, batch 3900, loss[loss=0.3446, simple_loss=0.3988, pruned_loss=0.1452, over 8604.00 frames. ], tot_loss[loss=0.3126, simple_loss=0.366, pruned_loss=0.1296, over 1609634.60 frames. ], batch size: 34, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:24:42,324 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8341, 2.3218, 4.5763, 1.2493, 2.6954, 2.2139, 1.7803, 2.4410], + device='cuda:1'), covar=tensor([0.1239, 0.1470, 0.0616, 0.2723, 0.1465, 0.2027, 0.1243, 0.2225], + device='cuda:1'), in_proj_covar=tensor([0.0451, 0.0427, 0.0502, 0.0510, 0.0562, 0.0491, 0.0434, 0.0564], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:25:02,709 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28191.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:25:08,592 INFO [train.py:901] (1/4) Epoch 4, batch 3950, loss[loss=0.3078, simple_loss=0.3635, pruned_loss=0.1261, over 8528.00 frames. ], tot_loss[loss=0.3096, simple_loss=0.3641, pruned_loss=0.1276, over 1609090.31 frames. ], batch size: 26, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:25:24,838 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.029e+02 3.357e+02 4.080e+02 5.453e+02 1.389e+03, threshold=8.161e+02, percent-clipped=8.0 +2023-02-05 23:25:41,190 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28247.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:25:43,103 INFO [train.py:901] (1/4) Epoch 4, batch 4000, loss[loss=0.273, simple_loss=0.3366, pruned_loss=0.1047, over 6472.00 frames. ], tot_loss[loss=0.3084, simple_loss=0.3628, pruned_loss=0.127, over 1603275.52 frames. ], batch size: 14, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:25:59,936 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28273.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:26:04,985 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-05 23:26:17,618 INFO [train.py:901] (1/4) Epoch 4, batch 4050, loss[loss=0.3244, simple_loss=0.3865, pruned_loss=0.1311, over 8334.00 frames. ], tot_loss[loss=0.3071, simple_loss=0.3619, pruned_loss=0.1261, over 1604753.02 frames. ], batch size: 26, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:26:17,834 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8498, 2.2813, 3.0367, 1.2639, 3.0313, 2.0055, 1.6388, 2.0981], + device='cuda:1'), covar=tensor([0.0249, 0.0130, 0.0072, 0.0188, 0.0103, 0.0257, 0.0219, 0.0111], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0192, 0.0155, 0.0229, 0.0187, 0.0313, 0.0250, 0.0220], + device='cuda:1'), out_proj_covar=tensor([1.0753e-04, 7.7614e-05, 5.9638e-05, 9.0650e-05, 7.6804e-05, 1.3637e-04, + 1.0344e-04, 8.6985e-05], device='cuda:1') +2023-02-05 23:26:21,157 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28305.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:26:24,887 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-05 23:26:34,456 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.986e+02 3.482e+02 4.201e+02 5.400e+02 1.078e+03, threshold=8.403e+02, percent-clipped=4.0 +2023-02-05 23:26:40,835 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.30 vs. limit=5.0 +2023-02-05 23:26:52,367 INFO [train.py:901] (1/4) Epoch 4, batch 4100, loss[loss=0.2799, simple_loss=0.3551, pruned_loss=0.1024, over 8459.00 frames. ], tot_loss[loss=0.3086, simple_loss=0.3631, pruned_loss=0.127, over 1608904.56 frames. ], batch size: 25, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:26:53,951 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3595, 2.2037, 1.5658, 1.9383, 1.8724, 1.3300, 1.7987, 1.8063], + device='cuda:1'), covar=tensor([0.0985, 0.0330, 0.0796, 0.0420, 0.0592, 0.1013, 0.0711, 0.0563], + device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0239, 0.0311, 0.0307, 0.0334, 0.0317, 0.0342, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 23:27:09,280 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 +2023-02-05 23:27:27,345 INFO [train.py:901] (1/4) Epoch 4, batch 4150, loss[loss=0.4314, simple_loss=0.4485, pruned_loss=0.2071, over 6785.00 frames. ], tot_loss[loss=0.3091, simple_loss=0.3636, pruned_loss=0.1273, over 1607622.21 frames. ], batch size: 71, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:27:43,612 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.222e+02 3.372e+02 4.170e+02 5.520e+02 1.384e+03, threshold=8.341e+02, percent-clipped=6.0 +2023-02-05 23:28:00,681 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28447.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:28:02,476 INFO [train.py:901] (1/4) Epoch 4, batch 4200, loss[loss=0.2369, simple_loss=0.2883, pruned_loss=0.09275, over 7270.00 frames. ], tot_loss[loss=0.3086, simple_loss=0.3633, pruned_loss=0.127, over 1604184.64 frames. ], batch size: 16, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:28:07,668 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-05 23:28:17,495 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28472.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:28:29,059 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-05 23:28:29,825 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7610, 5.8483, 5.1186, 2.3640, 5.2844, 5.3510, 5.4202, 4.5915], + device='cuda:1'), covar=tensor([0.0568, 0.0343, 0.0715, 0.3956, 0.0545, 0.0498, 0.0930, 0.0622], + device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0269, 0.0301, 0.0394, 0.0297, 0.0243, 0.0289, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 23:28:36,446 INFO [train.py:901] (1/4) Epoch 4, batch 4250, loss[loss=0.3604, simple_loss=0.4033, pruned_loss=0.1587, over 8526.00 frames. ], tot_loss[loss=0.3096, simple_loss=0.3644, pruned_loss=0.1274, over 1608756.74 frames. ], batch size: 49, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:28:39,187 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28504.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:28:43,323 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28510.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:28:51,867 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 3.170e+02 4.105e+02 5.662e+02 1.430e+03, threshold=8.210e+02, percent-clipped=9.0 +2023-02-05 23:29:07,173 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6589, 2.4909, 2.9950, 0.9556, 2.9401, 1.8869, 1.3292, 1.7192], + device='cuda:1'), covar=tensor([0.0225, 0.0088, 0.0057, 0.0239, 0.0115, 0.0245, 0.0290, 0.0132], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0192, 0.0155, 0.0234, 0.0185, 0.0314, 0.0253, 0.0220], + device='cuda:1'), out_proj_covar=tensor([1.1074e-04, 7.7224e-05, 5.9519e-05, 9.2067e-05, 7.6167e-05, 1.3641e-04, + 1.0452e-04, 8.6385e-05], device='cuda:1') +2023-02-05 23:29:10,384 INFO [train.py:901] (1/4) Epoch 4, batch 4300, loss[loss=0.3391, simple_loss=0.396, pruned_loss=0.1411, over 8617.00 frames. ], tot_loss[loss=0.3077, simple_loss=0.3635, pruned_loss=0.1259, over 1611034.71 frames. ], batch size: 34, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:29:37,893 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2213, 2.2034, 1.5595, 2.0144, 1.7083, 1.3360, 1.7212, 1.8747], + device='cuda:1'), covar=tensor([0.0937, 0.0377, 0.0918, 0.0426, 0.0654, 0.1128, 0.0730, 0.0599], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0238, 0.0307, 0.0311, 0.0326, 0.0310, 0.0342, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 23:29:38,454 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28591.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:29:45,231 INFO [train.py:901] (1/4) Epoch 4, batch 4350, loss[loss=0.2569, simple_loss=0.3227, pruned_loss=0.09555, over 7703.00 frames. ], tot_loss[loss=0.3083, simple_loss=0.3639, pruned_loss=0.1264, over 1612304.09 frames. ], batch size: 18, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:29:57,578 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28617.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:29:58,771 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-05 23:30:01,436 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.096e+02 3.285e+02 3.917e+02 4.771e+02 1.131e+03, threshold=7.833e+02, percent-clipped=1.0 +2023-02-05 23:30:19,075 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28649.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:30:19,684 INFO [train.py:901] (1/4) Epoch 4, batch 4400, loss[loss=0.2934, simple_loss=0.3528, pruned_loss=0.117, over 8141.00 frames. ], tot_loss[loss=0.3075, simple_loss=0.3634, pruned_loss=0.1257, over 1614169.30 frames. ], batch size: 22, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:30:23,359 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-05 23:30:36,843 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.38 vs. limit=5.0 +2023-02-05 23:30:41,085 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-05 23:30:54,265 INFO [train.py:901] (1/4) Epoch 4, batch 4450, loss[loss=0.2962, simple_loss=0.348, pruned_loss=0.1222, over 7924.00 frames. ], tot_loss[loss=0.3073, simple_loss=0.3634, pruned_loss=0.1256, over 1617138.17 frames. ], batch size: 20, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:30:58,500 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28706.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:31:10,735 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.181e+02 3.229e+02 4.056e+02 4.786e+02 8.259e+02, threshold=8.113e+02, percent-clipped=1.0 +2023-02-05 23:31:17,749 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28732.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:31:21,627 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.27 vs. limit=5.0 +2023-02-05 23:31:30,203 INFO [train.py:901] (1/4) Epoch 4, batch 4500, loss[loss=0.2782, simple_loss=0.3183, pruned_loss=0.119, over 7207.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.3619, pruned_loss=0.1246, over 1612717.41 frames. ], batch size: 16, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:31:36,227 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-05 23:31:39,889 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28764.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:32:05,460 INFO [train.py:901] (1/4) Epoch 4, batch 4550, loss[loss=0.2694, simple_loss=0.326, pruned_loss=0.1064, over 7451.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.3621, pruned_loss=0.1246, over 1616722.88 frames. ], batch size: 17, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:32:21,345 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 3.133e+02 4.046e+02 5.517e+02 1.256e+03, threshold=8.093e+02, percent-clipped=3.0 +2023-02-05 23:32:39,631 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:32:40,904 INFO [train.py:901] (1/4) Epoch 4, batch 4600, loss[loss=0.2758, simple_loss=0.334, pruned_loss=0.1088, over 8069.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.3619, pruned_loss=0.1247, over 1613421.59 frames. ], batch size: 21, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:32:43,600 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28854.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:33:00,433 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28879.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:33:13,243 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.25 vs. limit=5.0 +2023-02-05 23:33:14,780 INFO [train.py:901] (1/4) Epoch 4, batch 4650, loss[loss=0.2413, simple_loss=0.3084, pruned_loss=0.08714, over 7815.00 frames. ], tot_loss[loss=0.3081, simple_loss=0.3639, pruned_loss=0.1262, over 1618922.10 frames. ], batch size: 20, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:33:21,519 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3832, 1.4527, 1.6023, 1.3086, 0.8651, 1.6013, 0.1205, 1.1211], + device='cuda:1'), covar=tensor([0.3469, 0.2573, 0.1182, 0.1987, 0.6998, 0.1092, 0.5756, 0.2382], + device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0111, 0.0080, 0.0154, 0.0198, 0.0083, 0.0142, 0.0117], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 23:33:30,676 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 3.425e+02 4.570e+02 5.631e+02 1.457e+03, threshold=9.141e+02, percent-clipped=7.0 +2023-02-05 23:33:49,339 INFO [train.py:901] (1/4) Epoch 4, batch 4700, loss[loss=0.2828, simple_loss=0.3499, pruned_loss=0.1078, over 8464.00 frames. ], tot_loss[loss=0.309, simple_loss=0.3641, pruned_loss=0.1269, over 1616907.01 frames. ], batch size: 25, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:33:54,997 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6601, 4.6250, 4.0819, 1.7971, 4.0229, 4.1516, 4.2848, 3.8443], + device='cuda:1'), covar=tensor([0.0748, 0.0644, 0.0843, 0.4762, 0.0748, 0.0821, 0.1212, 0.0633], + device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0263, 0.0291, 0.0382, 0.0291, 0.0238, 0.0281, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 23:33:58,473 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28962.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:33:59,162 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28963.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:03,894 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28969.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:15,878 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28987.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:16,552 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28988.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:34:24,389 INFO [train.py:901] (1/4) Epoch 4, batch 4750, loss[loss=0.2533, simple_loss=0.3073, pruned_loss=0.09963, over 7927.00 frames. ], tot_loss[loss=0.3084, simple_loss=0.3638, pruned_loss=0.1265, over 1618845.30 frames. ], batch size: 20, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:34:33,279 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29013.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:34:38,666 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29020.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:38,747 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29020.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:40,432 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.078e+02 3.145e+02 3.754e+02 5.040e+02 8.107e+02, threshold=7.508e+02, percent-clipped=0.0 +2023-02-05 23:34:40,462 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-05 23:34:42,472 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-05 23:34:43,230 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6818, 5.6440, 5.0877, 1.6065, 5.0433, 5.2272, 5.2289, 4.5944], + device='cuda:1'), covar=tensor([0.0650, 0.0372, 0.0757, 0.4984, 0.0666, 0.0625, 0.1011, 0.0644], + device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0267, 0.0296, 0.0380, 0.0293, 0.0240, 0.0286, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 23:34:56,221 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29045.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:59,307 INFO [train.py:901] (1/4) Epoch 4, batch 4800, loss[loss=0.251, simple_loss=0.3161, pruned_loss=0.09298, over 7637.00 frames. ], tot_loss[loss=0.309, simple_loss=0.3644, pruned_loss=0.1267, over 1622517.87 frames. ], batch size: 19, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:35:34,007 INFO [train.py:901] (1/4) Epoch 4, batch 4850, loss[loss=0.4041, simple_loss=0.4254, pruned_loss=0.1914, over 8271.00 frames. ], tot_loss[loss=0.3107, simple_loss=0.3655, pruned_loss=0.1279, over 1622433.00 frames. ], batch size: 48, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:35:34,044 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-05 23:35:39,703 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6170, 2.8436, 1.8681, 2.3092, 2.2514, 1.3921, 1.7796, 2.2933], + device='cuda:1'), covar=tensor([0.1231, 0.0356, 0.0897, 0.0597, 0.0604, 0.1271, 0.1043, 0.0738], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0237, 0.0302, 0.0300, 0.0326, 0.0311, 0.0338, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 23:35:49,596 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.099e+02 3.374e+02 4.405e+02 6.016e+02 1.134e+03, threshold=8.810e+02, percent-clipped=7.0 +2023-02-05 23:36:08,349 INFO [train.py:901] (1/4) Epoch 4, batch 4900, loss[loss=0.2493, simple_loss=0.3071, pruned_loss=0.09576, over 7646.00 frames. ], tot_loss[loss=0.3095, simple_loss=0.3648, pruned_loss=0.1271, over 1622830.36 frames. ], batch size: 19, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:36:41,949 INFO [train.py:901] (1/4) Epoch 4, batch 4950, loss[loss=0.3152, simple_loss=0.3728, pruned_loss=0.1287, over 8571.00 frames. ], tot_loss[loss=0.3063, simple_loss=0.362, pruned_loss=0.1254, over 1620549.56 frames. ], batch size: 39, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:36:46,762 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2684, 4.0777, 3.7605, 1.4820, 3.7784, 3.5827, 3.8963, 3.4486], + device='cuda:1'), covar=tensor([0.0876, 0.0673, 0.1091, 0.4939, 0.0827, 0.0766, 0.1143, 0.0666], + device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0277, 0.0300, 0.0387, 0.0300, 0.0246, 0.0293, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-05 23:36:56,268 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29219.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:36:58,766 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.095e+02 3.208e+02 3.912e+02 5.596e+02 9.849e+02, threshold=7.824e+02, percent-clipped=2.0 +2023-02-05 23:36:58,865 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29223.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:00,341 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29225.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:01,726 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0403, 1.0464, 0.9419, 0.9823, 0.7208, 1.1481, 0.0368, 0.8566], + device='cuda:1'), covar=tensor([0.3270, 0.2010, 0.1168, 0.1837, 0.5015, 0.0931, 0.4705, 0.1967], + device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0107, 0.0080, 0.0151, 0.0191, 0.0082, 0.0136, 0.0116], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 23:37:12,969 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29244.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:16,643 INFO [train.py:901] (1/4) Epoch 4, batch 5000, loss[loss=0.2786, simple_loss=0.3327, pruned_loss=0.1122, over 7942.00 frames. ], tot_loss[loss=0.3058, simple_loss=0.3615, pruned_loss=0.125, over 1618926.33 frames. ], batch size: 20, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:37:16,867 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29250.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:19,511 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29254.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:34,853 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6190, 2.6963, 2.7269, 2.1786, 1.4249, 3.0019, 0.5178, 1.8716], + device='cuda:1'), covar=tensor([0.4669, 0.1546, 0.1218, 0.3599, 0.6591, 0.1046, 0.6742, 0.2552], + device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0109, 0.0083, 0.0156, 0.0192, 0.0082, 0.0140, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 23:37:51,640 INFO [train.py:901] (1/4) Epoch 4, batch 5050, loss[loss=0.2406, simple_loss=0.2999, pruned_loss=0.09064, over 7928.00 frames. ], tot_loss[loss=0.3064, simple_loss=0.3618, pruned_loss=0.1255, over 1616974.35 frames. ], batch size: 20, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:38:07,701 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 3.436e+02 4.072e+02 5.001e+02 1.022e+03, threshold=8.144e+02, percent-clipped=3.0 +2023-02-05 23:38:14,944 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-05 23:38:18,449 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29338.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:38:26,644 INFO [train.py:901] (1/4) Epoch 4, batch 5100, loss[loss=0.2402, simple_loss=0.304, pruned_loss=0.08816, over 7797.00 frames. ], tot_loss[loss=0.3049, simple_loss=0.3606, pruned_loss=0.1246, over 1618194.58 frames. ], batch size: 19, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:38:36,222 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29364.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:39:00,484 INFO [train.py:901] (1/4) Epoch 4, batch 5150, loss[loss=0.2447, simple_loss=0.3157, pruned_loss=0.0869, over 7648.00 frames. ], tot_loss[loss=0.303, simple_loss=0.3595, pruned_loss=0.1232, over 1614021.88 frames. ], batch size: 19, lr: 1.78e-02, grad_scale: 8.0 +2023-02-05 23:39:13,945 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2554, 1.4675, 2.3089, 1.0255, 1.5803, 1.4278, 1.2549, 1.4567], + device='cuda:1'), covar=tensor([0.1413, 0.1504, 0.0577, 0.2742, 0.1121, 0.2191, 0.1364, 0.1455], + device='cuda:1'), in_proj_covar=tensor([0.0455, 0.0433, 0.0513, 0.0517, 0.0559, 0.0501, 0.0440, 0.0571], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:39:16,241 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 3.145e+02 3.888e+02 4.871e+02 1.199e+03, threshold=7.777e+02, percent-clipped=1.0 +2023-02-05 23:39:35,364 INFO [train.py:901] (1/4) Epoch 4, batch 5200, loss[loss=0.311, simple_loss=0.3789, pruned_loss=0.1216, over 8494.00 frames. ], tot_loss[loss=0.3031, simple_loss=0.3594, pruned_loss=0.1234, over 1609442.26 frames. ], batch size: 28, lr: 1.78e-02, grad_scale: 8.0 +2023-02-05 23:39:37,871 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-05 23:39:54,954 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29479.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:40:09,486 INFO [train.py:901] (1/4) Epoch 4, batch 5250, loss[loss=0.2881, simple_loss=0.3551, pruned_loss=0.1106, over 8244.00 frames. ], tot_loss[loss=0.302, simple_loss=0.359, pruned_loss=0.1225, over 1611120.01 frames. ], batch size: 22, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:40:12,194 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-05 23:40:25,985 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.346e+02 3.507e+02 4.371e+02 5.555e+02 1.318e+03, threshold=8.742e+02, percent-clipped=11.0 +2023-02-05 23:40:43,420 INFO [train.py:901] (1/4) Epoch 4, batch 5300, loss[loss=0.308, simple_loss=0.3516, pruned_loss=0.1322, over 7814.00 frames. ], tot_loss[loss=0.3025, simple_loss=0.3591, pruned_loss=0.123, over 1611375.42 frames. ], batch size: 20, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:40:57,665 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29569.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:41:14,734 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29594.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:41:16,736 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5962, 1.1426, 1.2669, 1.0105, 1.1641, 1.1317, 1.3215, 1.2150], + device='cuda:1'), covar=tensor([0.0675, 0.1462, 0.2009, 0.1605, 0.0651, 0.1798, 0.0794, 0.0669], + device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0195, 0.0230, 0.0195, 0.0152, 0.0199, 0.0161, 0.0162], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0007, 0.0007, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:1') +2023-02-05 23:41:17,286 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29598.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:41:18,614 INFO [train.py:901] (1/4) Epoch 4, batch 5350, loss[loss=0.2933, simple_loss=0.3609, pruned_loss=0.1128, over 8136.00 frames. ], tot_loss[loss=0.3031, simple_loss=0.3596, pruned_loss=0.1233, over 1611334.57 frames. ], batch size: 22, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:41:32,358 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29619.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:41:35,507 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.071e+02 3.127e+02 4.006e+02 4.952e+02 2.682e+03, threshold=8.012e+02, percent-clipped=7.0 +2023-02-05 23:41:53,621 INFO [train.py:901] (1/4) Epoch 4, batch 5400, loss[loss=0.2702, simple_loss=0.3246, pruned_loss=0.1079, over 7529.00 frames. ], tot_loss[loss=0.3035, simple_loss=0.3596, pruned_loss=0.1236, over 1609569.42 frames. ], batch size: 18, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:42:14,529 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29680.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:42:28,600 INFO [train.py:901] (1/4) Epoch 4, batch 5450, loss[loss=0.2656, simple_loss=0.3301, pruned_loss=0.1005, over 8142.00 frames. ], tot_loss[loss=0.3045, simple_loss=0.3606, pruned_loss=0.1242, over 1606683.64 frames. ], batch size: 22, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:42:37,302 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29713.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:42:44,948 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.180e+02 3.089e+02 4.007e+02 5.016e+02 9.074e+02, threshold=8.014e+02, percent-clipped=4.0 +2023-02-05 23:42:52,684 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29735.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:42:57,993 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-05 23:43:02,828 INFO [train.py:901] (1/4) Epoch 4, batch 5500, loss[loss=0.3171, simple_loss=0.3715, pruned_loss=0.1313, over 8294.00 frames. ], tot_loss[loss=0.304, simple_loss=0.3607, pruned_loss=0.1237, over 1604936.10 frames. ], batch size: 23, lr: 1.77e-02, grad_scale: 4.0 +2023-02-05 23:43:09,623 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2757, 1.4536, 1.1724, 1.9290, 0.8966, 1.1536, 1.2643, 1.5154], + device='cuda:1'), covar=tensor([0.1404, 0.1257, 0.2034, 0.0774, 0.1623, 0.2303, 0.1351, 0.1070], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0282, 0.0297, 0.0221, 0.0264, 0.0292, 0.0302, 0.0274], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-05 23:43:10,332 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29760.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:43:23,276 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-02-05 23:43:38,374 INFO [train.py:901] (1/4) Epoch 4, batch 5550, loss[loss=0.3431, simple_loss=0.3902, pruned_loss=0.1481, over 8337.00 frames. ], tot_loss[loss=0.3053, simple_loss=0.3615, pruned_loss=0.1245, over 1605613.15 frames. ], batch size: 25, lr: 1.77e-02, grad_scale: 4.0 +2023-02-05 23:43:51,763 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29820.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:43:54,237 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 3.211e+02 3.931e+02 4.808e+02 9.688e+02, threshold=7.861e+02, percent-clipped=2.0 +2023-02-05 23:44:12,162 INFO [train.py:901] (1/4) Epoch 4, batch 5600, loss[loss=0.3268, simple_loss=0.3846, pruned_loss=0.1345, over 8329.00 frames. ], tot_loss[loss=0.3033, simple_loss=0.3604, pruned_loss=0.1231, over 1608915.42 frames. ], batch size: 25, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:44:46,063 INFO [train.py:901] (1/4) Epoch 4, batch 5650, loss[loss=0.3325, simple_loss=0.3751, pruned_loss=0.1449, over 8134.00 frames. ], tot_loss[loss=0.3054, simple_loss=0.3618, pruned_loss=0.1245, over 1612909.76 frames. ], batch size: 22, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:44:55,378 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29913.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:45:03,290 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 3.236e+02 4.025e+02 5.119e+02 8.732e+02, threshold=8.050e+02, percent-clipped=2.0 +2023-02-05 23:45:03,322 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-05 23:45:20,777 INFO [train.py:901] (1/4) Epoch 4, batch 5700, loss[loss=0.2502, simple_loss=0.3368, pruned_loss=0.0818, over 8337.00 frames. ], tot_loss[loss=0.3055, simple_loss=0.3621, pruned_loss=0.1245, over 1614481.20 frames. ], batch size: 25, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:45:34,475 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29969.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:45:41,359 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-02-05 23:45:48,656 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8634, 2.2484, 1.7073, 2.5522, 1.2151, 1.3006, 1.7539, 2.3459], + device='cuda:1'), covar=tensor([0.1159, 0.1133, 0.1528, 0.0540, 0.1659, 0.2254, 0.1539, 0.1027], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0287, 0.0301, 0.0224, 0.0264, 0.0296, 0.0310, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-05 23:45:52,052 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29994.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:45:55,961 INFO [train.py:901] (1/4) Epoch 4, batch 5750, loss[loss=0.3212, simple_loss=0.3869, pruned_loss=0.1277, over 8512.00 frames. ], tot_loss[loss=0.3046, simple_loss=0.3612, pruned_loss=0.124, over 1616416.33 frames. ], batch size: 26, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:46:07,174 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-05 23:46:13,263 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.352e+02 3.278e+02 4.024e+02 4.787e+02 1.009e+03, threshold=8.047e+02, percent-clipped=4.0 +2023-02-05 23:46:13,350 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30024.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:46:17,493 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30028.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:46:32,295 INFO [train.py:901] (1/4) Epoch 4, batch 5800, loss[loss=0.3638, simple_loss=0.3941, pruned_loss=0.1667, over 6946.00 frames. ], tot_loss[loss=0.3059, simple_loss=0.3619, pruned_loss=0.125, over 1615469.88 frames. ], batch size: 71, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:47:06,538 INFO [train.py:901] (1/4) Epoch 4, batch 5850, loss[loss=0.3304, simple_loss=0.3806, pruned_loss=0.1401, over 8143.00 frames. ], tot_loss[loss=0.3074, simple_loss=0.3629, pruned_loss=0.126, over 1614115.28 frames. ], batch size: 22, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:47:23,103 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 3.427e+02 4.657e+02 5.932e+02 9.223e+02, threshold=9.314e+02, percent-clipped=4.0 +2023-02-05 23:47:30,861 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0568, 2.2540, 4.1156, 3.1132, 3.1298, 2.3649, 1.6373, 1.6911], + device='cuda:1'), covar=tensor([0.1338, 0.1934, 0.0348, 0.0827, 0.0848, 0.0740, 0.0886, 0.1791], + device='cuda:1'), in_proj_covar=tensor([0.0687, 0.0615, 0.0522, 0.0589, 0.0701, 0.0576, 0.0559, 0.0574], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-05 23:47:33,291 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30139.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:47:35,877 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30143.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:47:41,683 INFO [train.py:901] (1/4) Epoch 4, batch 5900, loss[loss=0.274, simple_loss=0.3481, pruned_loss=0.09995, over 8246.00 frames. ], tot_loss[loss=0.307, simple_loss=0.3624, pruned_loss=0.1258, over 1613022.04 frames. ], batch size: 22, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:47:51,359 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30164.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:48:15,969 INFO [train.py:901] (1/4) Epoch 4, batch 5950, loss[loss=0.2981, simple_loss=0.3522, pruned_loss=0.122, over 8235.00 frames. ], tot_loss[loss=0.3073, simple_loss=0.3623, pruned_loss=0.1261, over 1611976.97 frames. ], batch size: 22, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:48:32,439 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 3.143e+02 3.968e+02 4.977e+02 1.070e+03, threshold=7.937e+02, percent-clipped=1.0 +2023-02-05 23:48:35,345 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30227.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:48:50,186 INFO [train.py:901] (1/4) Epoch 4, batch 6000, loss[loss=0.3083, simple_loss=0.3693, pruned_loss=0.1237, over 8612.00 frames. ], tot_loss[loss=0.3077, simple_loss=0.3625, pruned_loss=0.1264, over 1613370.28 frames. ], batch size: 31, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:48:50,186 INFO [train.py:926] (1/4) Computing validation loss +2023-02-05 23:49:02,859 INFO [train.py:935] (1/4) Epoch 4, validation: loss=0.2338, simple_loss=0.3275, pruned_loss=0.07005, over 944034.00 frames. +2023-02-05 23:49:02,860 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-05 23:49:22,554 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30279.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:49:25,895 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30284.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:49:37,706 INFO [train.py:901] (1/4) Epoch 4, batch 6050, loss[loss=0.2916, simple_loss=0.3571, pruned_loss=0.113, over 8288.00 frames. ], tot_loss[loss=0.3072, simple_loss=0.3622, pruned_loss=0.126, over 1609751.30 frames. ], batch size: 23, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:49:44,061 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30309.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:49:53,935 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 3.338e+02 3.992e+02 4.649e+02 1.183e+03, threshold=7.984e+02, percent-clipped=3.0 +2023-02-05 23:50:12,470 INFO [train.py:901] (1/4) Epoch 4, batch 6100, loss[loss=0.3137, simple_loss=0.3609, pruned_loss=0.1333, over 7800.00 frames. ], tot_loss[loss=0.3067, simple_loss=0.3619, pruned_loss=0.1257, over 1610271.07 frames. ], batch size: 19, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:50:32,433 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30378.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:50:39,279 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-05 23:50:44,138 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30395.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:50:47,304 INFO [train.py:901] (1/4) Epoch 4, batch 6150, loss[loss=0.381, simple_loss=0.4265, pruned_loss=0.1678, over 8472.00 frames. ], tot_loss[loss=0.3065, simple_loss=0.3618, pruned_loss=0.1256, over 1613012.35 frames. ], batch size: 29, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:50:51,617 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6703, 2.3594, 3.0513, 0.7183, 3.0465, 1.8167, 1.3302, 1.5327], + device='cuda:1'), covar=tensor([0.0342, 0.0109, 0.0079, 0.0291, 0.0163, 0.0310, 0.0361, 0.0180], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0196, 0.0163, 0.0241, 0.0190, 0.0323, 0.0259, 0.0223], + device='cuda:1'), out_proj_covar=tensor([1.0953e-04, 7.6379e-05, 6.0923e-05, 9.1732e-05, 7.5420e-05, 1.3652e-04, + 1.0317e-04, 8.6125e-05], device='cuda:1') +2023-02-05 23:50:57,627 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2312, 1.8633, 1.2125, 1.6262, 1.4752, 1.1253, 1.3988, 1.5586], + device='cuda:1'), covar=tensor([0.0698, 0.0237, 0.0663, 0.0349, 0.0439, 0.0762, 0.0574, 0.0486], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0235, 0.0307, 0.0304, 0.0332, 0.0314, 0.0337, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 23:51:02,475 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30420.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:51:05,069 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 3.511e+02 4.267e+02 5.249e+02 1.089e+03, threshold=8.535e+02, percent-clipped=6.0 +2023-02-05 23:51:23,145 INFO [train.py:901] (1/4) Epoch 4, batch 6200, loss[loss=0.2899, simple_loss=0.3593, pruned_loss=0.1102, over 8247.00 frames. ], tot_loss[loss=0.3076, simple_loss=0.3628, pruned_loss=0.1262, over 1615923.78 frames. ], batch size: 24, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:51:36,029 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6722, 2.8513, 1.6301, 2.0854, 2.1350, 1.4360, 1.7640, 2.0439], + device='cuda:1'), covar=tensor([0.1373, 0.0291, 0.1012, 0.0769, 0.0750, 0.1323, 0.1108, 0.0897], + device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0243, 0.0313, 0.0316, 0.0339, 0.0320, 0.0342, 0.0319], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-05 23:51:48,270 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30487.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:51:57,470 INFO [train.py:901] (1/4) Epoch 4, batch 6250, loss[loss=0.2639, simple_loss=0.3364, pruned_loss=0.09575, over 8094.00 frames. ], tot_loss[loss=0.309, simple_loss=0.3638, pruned_loss=0.1271, over 1617662.24 frames. ], batch size: 23, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:52:14,465 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 3.291e+02 3.933e+02 5.014e+02 1.132e+03, threshold=7.866e+02, percent-clipped=4.0 +2023-02-05 23:52:22,914 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30535.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:52:24,418 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-05 23:52:32,705 INFO [train.py:901] (1/4) Epoch 4, batch 6300, loss[loss=0.3623, simple_loss=0.4106, pruned_loss=0.157, over 8352.00 frames. ], tot_loss[loss=0.3085, simple_loss=0.3637, pruned_loss=0.1267, over 1614457.29 frames. ], batch size: 26, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:52:39,495 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30560.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:52:47,316 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30571.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:53:06,769 INFO [train.py:901] (1/4) Epoch 4, batch 6350, loss[loss=0.3402, simple_loss=0.4067, pruned_loss=0.1368, over 8501.00 frames. ], tot_loss[loss=0.3089, simple_loss=0.364, pruned_loss=0.1269, over 1615366.66 frames. ], batch size: 28, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:53:08,358 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30602.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:53:23,787 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 3.165e+02 3.849e+02 5.077e+02 1.430e+03, threshold=7.697e+02, percent-clipped=4.0 +2023-02-05 23:53:30,819 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-02-05 23:53:42,468 INFO [train.py:901] (1/4) Epoch 4, batch 6400, loss[loss=0.2566, simple_loss=0.324, pruned_loss=0.09463, over 8073.00 frames. ], tot_loss[loss=0.306, simple_loss=0.3616, pruned_loss=0.1252, over 1612130.76 frames. ], batch size: 21, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:53:43,310 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5506, 2.4717, 4.6761, 1.2283, 3.1249, 2.2409, 1.7768, 2.4221], + device='cuda:1'), covar=tensor([0.1418, 0.1415, 0.0527, 0.2802, 0.1260, 0.2008, 0.1281, 0.2167], + device='cuda:1'), in_proj_covar=tensor([0.0457, 0.0427, 0.0509, 0.0518, 0.0564, 0.0499, 0.0440, 0.0576], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:53:56,547 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-02-05 23:53:58,063 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30673.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:54:07,683 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30686.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:54:14,455 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9933, 2.3107, 3.7763, 1.5059, 2.8168, 2.3012, 2.0101, 2.5083], + device='cuda:1'), covar=tensor([0.0987, 0.1363, 0.0437, 0.2291, 0.0979, 0.1530, 0.0994, 0.1692], + device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0424, 0.0501, 0.0510, 0.0555, 0.0493, 0.0433, 0.0566], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:54:16,882 INFO [train.py:901] (1/4) Epoch 4, batch 6450, loss[loss=0.2964, simple_loss=0.359, pruned_loss=0.1169, over 8652.00 frames. ], tot_loss[loss=0.304, simple_loss=0.361, pruned_loss=0.1235, over 1615018.22 frames. ], batch size: 34, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:54:31,539 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30722.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:54:32,828 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.794e+02 3.022e+02 3.987e+02 5.645e+02 1.412e+03, threshold=7.975e+02, percent-clipped=10.0 +2023-02-05 23:54:50,958 INFO [train.py:901] (1/4) Epoch 4, batch 6500, loss[loss=0.2942, simple_loss=0.3627, pruned_loss=0.1128, over 8518.00 frames. ], tot_loss[loss=0.3036, simple_loss=0.3604, pruned_loss=0.1234, over 1616164.32 frames. ], batch size: 28, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:55:26,192 INFO [train.py:901] (1/4) Epoch 4, batch 6550, loss[loss=0.2442, simple_loss=0.306, pruned_loss=0.09118, over 6384.00 frames. ], tot_loss[loss=0.3027, simple_loss=0.3605, pruned_loss=0.1225, over 1615137.17 frames. ], batch size: 14, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:55:42,638 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.464e+02 3.539e+02 4.251e+02 5.114e+02 1.135e+03, threshold=8.501e+02, percent-clipped=1.0 +2023-02-05 23:55:50,034 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-05 23:55:51,488 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30837.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:56:00,640 INFO [train.py:901] (1/4) Epoch 4, batch 6600, loss[loss=0.2349, simple_loss=0.2977, pruned_loss=0.08607, over 7695.00 frames. ], tot_loss[loss=0.305, simple_loss=0.3626, pruned_loss=0.1237, over 1618907.84 frames. ], batch size: 18, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:56:06,240 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30858.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:56:08,699 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-05 23:56:24,234 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30883.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:56:35,343 INFO [train.py:901] (1/4) Epoch 4, batch 6650, loss[loss=0.321, simple_loss=0.3633, pruned_loss=0.1394, over 6885.00 frames. ], tot_loss[loss=0.3052, simple_loss=0.3624, pruned_loss=0.124, over 1621958.85 frames. ], batch size: 71, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:56:50,095 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30921.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:56:51,878 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.853e+02 3.362e+02 4.352e+02 5.461e+02 1.446e+03, threshold=8.703e+02, percent-clipped=3.0 +2023-02-05 23:57:01,675 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 +2023-02-05 23:57:04,055 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30942.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:57:04,665 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30943.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:57:09,171 INFO [train.py:901] (1/4) Epoch 4, batch 6700, loss[loss=0.3273, simple_loss=0.3461, pruned_loss=0.1542, over 7692.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.3621, pruned_loss=0.1245, over 1618667.31 frames. ], batch size: 18, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:57:10,676 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30952.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:57:21,616 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30967.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:57:45,057 INFO [train.py:901] (1/4) Epoch 4, batch 6750, loss[loss=0.2957, simple_loss=0.3602, pruned_loss=0.1156, over 8549.00 frames. ], tot_loss[loss=0.3035, simple_loss=0.3606, pruned_loss=0.1232, over 1620828.68 frames. ], batch size: 31, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:57:56,392 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31017.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:58:00,825 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.225e+02 3.317e+02 4.136e+02 5.252e+02 1.678e+03, threshold=8.272e+02, percent-clipped=4.0 +2023-02-05 23:58:04,859 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31029.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:58:07,647 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3994, 1.6910, 2.3870, 1.2210, 1.8533, 1.6427, 1.4252, 1.5281], + device='cuda:1'), covar=tensor([0.1341, 0.1493, 0.0583, 0.2565, 0.0986, 0.2032, 0.1293, 0.1399], + device='cuda:1'), in_proj_covar=tensor([0.0460, 0.0431, 0.0520, 0.0526, 0.0569, 0.0508, 0.0448, 0.0582], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-05 23:58:15,267 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.21 vs. limit=5.0 +2023-02-05 23:58:18,941 INFO [train.py:901] (1/4) Epoch 4, batch 6800, loss[loss=0.2853, simple_loss=0.3478, pruned_loss=0.1114, over 8092.00 frames. ], tot_loss[loss=0.3043, simple_loss=0.3612, pruned_loss=0.1237, over 1615336.51 frames. ], batch size: 21, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:58:19,604 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-05 23:58:48,702 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31093.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:58:53,203 INFO [train.py:901] (1/4) Epoch 4, batch 6850, loss[loss=0.3699, simple_loss=0.408, pruned_loss=0.1659, over 8517.00 frames. ], tot_loss[loss=0.3029, simple_loss=0.3595, pruned_loss=0.1231, over 1610408.27 frames. ], batch size: 49, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:59:06,887 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31118.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:59:10,004 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-05 23:59:10,569 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 3.178e+02 3.797e+02 5.313e+02 1.260e+03, threshold=7.594e+02, percent-clipped=4.0 +2023-02-05 23:59:16,196 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31132.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:59:28,759 INFO [train.py:901] (1/4) Epoch 4, batch 6900, loss[loss=0.2881, simple_loss=0.363, pruned_loss=0.1066, over 8324.00 frames. ], tot_loss[loss=0.3017, simple_loss=0.3589, pruned_loss=0.1223, over 1614976.07 frames. ], batch size: 25, lr: 1.73e-02, grad_scale: 8.0 +2023-02-05 23:59:28,928 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5798, 1.7140, 3.2544, 1.2411, 2.2371, 3.5235, 3.2663, 2.9823], + device='cuda:1'), covar=tensor([0.1262, 0.1331, 0.0348, 0.2152, 0.0768, 0.0262, 0.0392, 0.0616], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0257, 0.0214, 0.0259, 0.0211, 0.0194, 0.0204, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 00:00:03,318 INFO [train.py:901] (1/4) Epoch 4, batch 6950, loss[loss=0.3282, simple_loss=0.3862, pruned_loss=0.1351, over 8479.00 frames. ], tot_loss[loss=0.3029, simple_loss=0.3598, pruned_loss=0.123, over 1614676.89 frames. ], batch size: 25, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:00:06,208 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2760, 1.5842, 1.4780, 1.2231, 1.6526, 1.4631, 1.7966, 1.7889], + device='cuda:1'), covar=tensor([0.0644, 0.1345, 0.2065, 0.1669, 0.0622, 0.1690, 0.0804, 0.0576], + device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0193, 0.0235, 0.0195, 0.0148, 0.0198, 0.0160, 0.0159], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:00:18,141 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 00:00:20,074 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 3.425e+02 4.122e+02 5.302e+02 9.579e+02, threshold=8.244e+02, percent-clipped=6.0 +2023-02-06 00:00:38,131 INFO [train.py:901] (1/4) Epoch 4, batch 7000, loss[loss=0.3544, simple_loss=0.3996, pruned_loss=0.1546, over 8477.00 frames. ], tot_loss[loss=0.305, simple_loss=0.3608, pruned_loss=0.1246, over 1609553.26 frames. ], batch size: 27, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:00:48,755 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31265.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:00:54,169 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8741, 1.3721, 5.8954, 2.0537, 5.2701, 4.9628, 5.4987, 5.4624], + device='cuda:1'), covar=tensor([0.0457, 0.3494, 0.0225, 0.2112, 0.0880, 0.0518, 0.0387, 0.0381], + device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0465, 0.0367, 0.0386, 0.0454, 0.0377, 0.0374, 0.0415], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 00:01:03,270 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31287.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:01:09,178 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31296.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:01:11,662 INFO [train.py:901] (1/4) Epoch 4, batch 7050, loss[loss=0.2332, simple_loss=0.2947, pruned_loss=0.08582, over 7699.00 frames. ], tot_loss[loss=0.3036, simple_loss=0.3599, pruned_loss=0.1237, over 1612167.08 frames. ], batch size: 18, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:01:28,375 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 3.221e+02 3.873e+02 5.326e+02 1.178e+03, threshold=7.746e+02, percent-clipped=8.0 +2023-02-06 00:01:47,442 INFO [train.py:901] (1/4) Epoch 4, batch 7100, loss[loss=0.3002, simple_loss=0.3464, pruned_loss=0.127, over 7430.00 frames. ], tot_loss[loss=0.3036, simple_loss=0.3599, pruned_loss=0.1236, over 1605698.63 frames. ], batch size: 17, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:01:59,550 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2070, 1.5242, 2.2051, 1.0230, 1.6754, 1.3479, 1.2769, 1.4275], + device='cuda:1'), covar=tensor([0.1220, 0.1385, 0.0480, 0.2461, 0.1056, 0.1844, 0.1205, 0.1552], + device='cuda:1'), in_proj_covar=tensor([0.0470, 0.0436, 0.0520, 0.0527, 0.0572, 0.0516, 0.0449, 0.0582], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:02:02,610 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31373.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:02:08,041 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31380.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:02:13,376 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:02:21,056 INFO [train.py:901] (1/4) Epoch 4, batch 7150, loss[loss=0.3437, simple_loss=0.3878, pruned_loss=0.1498, over 8185.00 frames. ], tot_loss[loss=0.306, simple_loss=0.3619, pruned_loss=0.1251, over 1609898.76 frames. ], batch size: 23, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:02:22,714 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31402.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:02:28,563 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31411.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:02:29,914 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31413.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:02:37,019 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 3.187e+02 3.955e+02 5.000e+02 8.847e+02, threshold=7.910e+02, percent-clipped=2.0 +2023-02-06 00:02:37,260 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4586, 2.0343, 2.2779, 0.5564, 2.2600, 1.5990, 0.5688, 1.8010], + device='cuda:1'), covar=tensor([0.0182, 0.0076, 0.0081, 0.0204, 0.0143, 0.0257, 0.0281, 0.0086], + device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0204, 0.0166, 0.0247, 0.0201, 0.0337, 0.0274, 0.0236], + device='cuda:1'), out_proj_covar=tensor([1.1113e-04, 7.8551e-05, 6.1639e-05, 9.2134e-05, 7.8388e-05, 1.4029e-04, + 1.0728e-04, 9.0225e-05], device='cuda:1') +2023-02-06 00:02:39,202 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31427.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:02:56,011 INFO [train.py:901] (1/4) Epoch 4, batch 7200, loss[loss=0.275, simple_loss=0.3254, pruned_loss=0.1123, over 7407.00 frames. ], tot_loss[loss=0.3036, simple_loss=0.3602, pruned_loss=0.1235, over 1613540.21 frames. ], batch size: 17, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:03:02,368 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5838, 1.8819, 3.1648, 1.1062, 2.3824, 1.7920, 1.5810, 1.8661], + device='cuda:1'), covar=tensor([0.1465, 0.1778, 0.0590, 0.3038, 0.1240, 0.2326, 0.1328, 0.2158], + device='cuda:1'), in_proj_covar=tensor([0.0458, 0.0427, 0.0508, 0.0513, 0.0556, 0.0503, 0.0437, 0.0572], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:03:04,340 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8130, 2.2729, 1.9256, 2.8869, 1.5231, 1.4547, 1.9841, 2.6386], + device='cuda:1'), covar=tensor([0.1216, 0.1176, 0.1464, 0.0577, 0.1711, 0.2386, 0.1584, 0.0742], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0278, 0.0293, 0.0218, 0.0261, 0.0292, 0.0292, 0.0267], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:03:22,241 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31488.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:03:23,586 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9124, 1.2604, 4.4279, 2.0301, 2.4038, 4.9675, 4.7437, 4.5382], + device='cuda:1'), covar=tensor([0.1305, 0.1697, 0.0279, 0.1815, 0.0847, 0.0219, 0.0324, 0.0520], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0265, 0.0218, 0.0265, 0.0222, 0.0198, 0.0210, 0.0270], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:03:30,969 INFO [train.py:901] (1/4) Epoch 4, batch 7250, loss[loss=0.2421, simple_loss=0.3197, pruned_loss=0.08227, over 7653.00 frames. ], tot_loss[loss=0.3016, simple_loss=0.3587, pruned_loss=0.1222, over 1615225.92 frames. ], batch size: 19, lr: 1.73e-02, grad_scale: 16.0 +2023-02-06 00:03:37,362 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31509.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:03:47,340 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.140e+02 3.150e+02 3.858e+02 4.938e+02 9.845e+02, threshold=7.715e+02, percent-clipped=4.0 +2023-02-06 00:03:47,838 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-06 00:04:05,056 INFO [train.py:901] (1/4) Epoch 4, batch 7300, loss[loss=0.2595, simple_loss=0.3392, pruned_loss=0.08993, over 7931.00 frames. ], tot_loss[loss=0.3004, simple_loss=0.3576, pruned_loss=0.1217, over 1606338.15 frames. ], batch size: 20, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:04:37,950 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6472, 2.1058, 3.6331, 1.1240, 2.2964, 1.7772, 1.7533, 1.9294], + device='cuda:1'), covar=tensor([0.1708, 0.1981, 0.0688, 0.3406, 0.1697, 0.2559, 0.1556, 0.2513], + device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0434, 0.0507, 0.0515, 0.0555, 0.0506, 0.0442, 0.0577], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:04:40,425 INFO [train.py:901] (1/4) Epoch 4, batch 7350, loss[loss=0.3338, simple_loss=0.3925, pruned_loss=0.1375, over 8361.00 frames. ], tot_loss[loss=0.3024, simple_loss=0.3593, pruned_loss=0.1228, over 1602426.11 frames. ], batch size: 24, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:04:57,284 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.182e+02 2.774e+02 3.613e+02 4.483e+02 1.102e+03, threshold=7.227e+02, percent-clipped=2.0 +2023-02-06 00:04:59,973 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 00:05:05,401 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:14,083 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3019, 1.5192, 1.4914, 1.0829, 1.5942, 1.4058, 1.7663, 1.5798], + device='cuda:1'), covar=tensor([0.0606, 0.1225, 0.1874, 0.1585, 0.0609, 0.1550, 0.0776, 0.0616], + device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0191, 0.0233, 0.0193, 0.0146, 0.0196, 0.0157, 0.0159], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:05:14,564 INFO [train.py:901] (1/4) Epoch 4, batch 7400, loss[loss=0.2945, simple_loss=0.359, pruned_loss=0.115, over 8131.00 frames. ], tot_loss[loss=0.3024, simple_loss=0.3592, pruned_loss=0.1228, over 1602753.56 frames. ], batch size: 22, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:05:19,266 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 00:05:20,112 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31658.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:21,983 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:26,692 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:37,009 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31683.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:43,494 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31692.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:48,576 INFO [train.py:901] (1/4) Epoch 4, batch 7450, loss[loss=0.2754, simple_loss=0.3495, pruned_loss=0.1006, over 8244.00 frames. ], tot_loss[loss=0.3039, simple_loss=0.3606, pruned_loss=0.1237, over 1608394.27 frames. ], batch size: 24, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:05:58,011 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 00:06:04,225 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31722.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:06:05,435 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.933e+02 3.216e+02 3.933e+02 5.503e+02 1.387e+03, threshold=7.866e+02, percent-clipped=9.0 +2023-02-06 00:06:19,838 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31744.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:06:23,714 INFO [train.py:901] (1/4) Epoch 4, batch 7500, loss[loss=0.396, simple_loss=0.4126, pruned_loss=0.1897, over 6950.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.362, pruned_loss=0.1246, over 1612739.84 frames. ], batch size: 74, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:06:27,341 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9291, 3.8826, 2.3510, 2.3970, 2.6564, 1.8276, 2.2559, 2.8275], + device='cuda:1'), covar=tensor([0.1423, 0.0259, 0.0821, 0.0737, 0.0705, 0.1107, 0.1102, 0.1033], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0236, 0.0310, 0.0303, 0.0324, 0.0311, 0.0336, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 00:06:36,769 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31769.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:06:37,956 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31771.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:06:58,133 INFO [train.py:901] (1/4) Epoch 4, batch 7550, loss[loss=0.2734, simple_loss=0.3448, pruned_loss=0.101, over 8240.00 frames. ], tot_loss[loss=0.3031, simple_loss=0.3597, pruned_loss=0.1232, over 1609214.16 frames. ], batch size: 22, lr: 1.72e-02, grad_scale: 8.0 +2023-02-06 00:07:02,947 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31806.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:07:16,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.990e+02 2.842e+02 3.963e+02 5.244e+02 1.193e+03, threshold=7.926e+02, percent-clipped=8.0 +2023-02-06 00:07:27,214 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:07:32,194 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5047, 3.4311, 2.4115, 4.1846, 1.9369, 2.0545, 2.1801, 3.4878], + device='cuda:1'), covar=tensor([0.1182, 0.0982, 0.1906, 0.0328, 0.2017, 0.2422, 0.2106, 0.0899], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0277, 0.0299, 0.0225, 0.0263, 0.0288, 0.0293, 0.0270], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:07:34,061 INFO [train.py:901] (1/4) Epoch 4, batch 7600, loss[loss=0.2731, simple_loss=0.3219, pruned_loss=0.1122, over 7216.00 frames. ], tot_loss[loss=0.3014, simple_loss=0.358, pruned_loss=0.1224, over 1606657.53 frames. ], batch size: 16, lr: 1.72e-02, grad_scale: 8.0 +2023-02-06 00:07:36,148 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31853.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:07:58,617 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:08:05,425 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31896.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:08:07,987 INFO [train.py:901] (1/4) Epoch 4, batch 7650, loss[loss=0.2208, simple_loss=0.2938, pruned_loss=0.07387, over 7939.00 frames. ], tot_loss[loss=0.3007, simple_loss=0.3579, pruned_loss=0.1218, over 1604893.15 frames. ], batch size: 20, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:08:26,153 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.122e+02 3.181e+02 3.860e+02 4.828e+02 9.649e+02, threshold=7.720e+02, percent-clipped=2.0 +2023-02-06 00:08:31,512 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31933.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:08:40,609 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.56 vs. limit=5.0 +2023-02-06 00:08:43,637 INFO [train.py:901] (1/4) Epoch 4, batch 7700, loss[loss=0.3253, simple_loss=0.38, pruned_loss=0.1352, over 8507.00 frames. ], tot_loss[loss=0.3006, simple_loss=0.3579, pruned_loss=0.1217, over 1605919.01 frames. ], batch size: 49, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:08:53,640 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 00:08:55,985 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31968.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:08:56,021 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31968.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:09:06,796 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 00:09:17,470 INFO [train.py:901] (1/4) Epoch 4, batch 7750, loss[loss=0.2463, simple_loss=0.3202, pruned_loss=0.08625, over 7800.00 frames. ], tot_loss[loss=0.3022, simple_loss=0.3587, pruned_loss=0.1229, over 1605880.22 frames. ], batch size: 19, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:09:35,964 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.165e+02 3.163e+02 3.927e+02 5.355e+02 1.239e+03, threshold=7.853e+02, percent-clipped=4.0 +2023-02-06 00:09:44,297 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6319, 2.9439, 1.7334, 2.1293, 2.2465, 1.4648, 2.0795, 2.2146], + device='cuda:1'), covar=tensor([0.1224, 0.0322, 0.0876, 0.0652, 0.0695, 0.1186, 0.0848, 0.0919], + device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0241, 0.0317, 0.0309, 0.0328, 0.0314, 0.0341, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 00:09:53,604 INFO [train.py:901] (1/4) Epoch 4, batch 7800, loss[loss=0.271, simple_loss=0.3519, pruned_loss=0.09508, over 8288.00 frames. ], tot_loss[loss=0.3018, simple_loss=0.359, pruned_loss=0.1223, over 1609624.52 frames. ], batch size: 23, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:10:05,166 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:10:27,396 INFO [train.py:901] (1/4) Epoch 4, batch 7850, loss[loss=0.3699, simple_loss=0.4115, pruned_loss=0.1641, over 8332.00 frames. ], tot_loss[loss=0.3001, simple_loss=0.3579, pruned_loss=0.1212, over 1614577.25 frames. ], batch size: 26, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:10:43,937 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.460e+02 3.521e+02 4.480e+02 6.179e+02 1.308e+03, threshold=8.960e+02, percent-clipped=13.0 +2023-02-06 00:10:55,616 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32142.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:00,779 INFO [train.py:901] (1/4) Epoch 4, batch 7900, loss[loss=0.2447, simple_loss=0.3148, pruned_loss=0.08725, over 7912.00 frames. ], tot_loss[loss=0.3009, simple_loss=0.3582, pruned_loss=0.1218, over 1605001.73 frames. ], batch size: 20, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:11:00,852 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32150.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:12,543 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32167.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:21,566 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32181.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:24,161 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32185.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:34,317 INFO [train.py:901] (1/4) Epoch 4, batch 7950, loss[loss=0.2882, simple_loss=0.347, pruned_loss=0.1147, over 7932.00 frames. ], tot_loss[loss=0.301, simple_loss=0.3585, pruned_loss=0.1217, over 1606498.93 frames. ], batch size: 20, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:11:51,235 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32224.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:11:51,628 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.248e+02 3.536e+02 4.226e+02 5.315e+02 1.259e+03, threshold=8.452e+02, percent-clipped=4.0 +2023-02-06 00:11:53,748 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3616, 1.6228, 1.5806, 0.4831, 1.5706, 1.1949, 0.3072, 1.6225], + device='cuda:1'), covar=tensor([0.0159, 0.0102, 0.0095, 0.0182, 0.0117, 0.0331, 0.0270, 0.0078], + device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0203, 0.0164, 0.0251, 0.0199, 0.0331, 0.0269, 0.0232], + device='cuda:1'), out_proj_covar=tensor([1.0900e-04, 7.7012e-05, 6.0510e-05, 9.3008e-05, 7.6624e-05, 1.3534e-04, + 1.0371e-04, 8.7393e-05], device='cuda:1') +2023-02-06 00:12:01,778 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32240.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:08,272 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32249.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:12:08,749 INFO [train.py:901] (1/4) Epoch 4, batch 8000, loss[loss=0.3073, simple_loss=0.3547, pruned_loss=0.1299, over 7772.00 frames. ], tot_loss[loss=0.3011, simple_loss=0.3579, pruned_loss=0.1222, over 1604095.44 frames. ], batch size: 19, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:12:11,585 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6202, 1.5863, 1.6574, 1.3926, 1.6104, 1.7061, 1.8913, 1.6422], + device='cuda:1'), covar=tensor([0.0626, 0.1297, 0.1918, 0.1530, 0.0667, 0.1575, 0.0823, 0.0638], + device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0189, 0.0232, 0.0194, 0.0144, 0.0198, 0.0159, 0.0161], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:12:19,078 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32265.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:26,260 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1555, 1.6886, 1.2322, 1.5470, 1.4715, 1.1151, 1.2797, 1.5530], + device='cuda:1'), covar=tensor([0.0748, 0.0294, 0.0596, 0.0302, 0.0356, 0.0789, 0.0537, 0.0479], + device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0239, 0.0315, 0.0309, 0.0323, 0.0314, 0.0340, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 00:12:26,798 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32277.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:41,283 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6850, 2.0710, 1.7738, 2.6554, 1.2199, 1.4722, 1.6802, 2.2157], + device='cuda:1'), covar=tensor([0.1293, 0.1207, 0.1601, 0.0593, 0.1727, 0.2181, 0.1583, 0.1180], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0277, 0.0300, 0.0224, 0.0259, 0.0294, 0.0297, 0.0269], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:12:42,459 INFO [train.py:901] (1/4) Epoch 4, batch 8050, loss[loss=0.2551, simple_loss=0.3191, pruned_loss=0.0956, over 7543.00 frames. ], tot_loss[loss=0.3013, simple_loss=0.3571, pruned_loss=0.1228, over 1587984.05 frames. ], batch size: 18, lr: 1.70e-02, grad_scale: 8.0 +2023-02-06 00:12:42,649 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32300.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:50,708 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:58,916 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.697e+02 3.496e+02 4.220e+02 5.135e+02 1.064e+03, threshold=8.441e+02, percent-clipped=2.0 +2023-02-06 00:13:02,564 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 00:13:15,984 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 00:13:19,663 INFO [train.py:901] (1/4) Epoch 5, batch 0, loss[loss=0.351, simple_loss=0.3988, pruned_loss=0.1516, over 8495.00 frames. ], tot_loss[loss=0.351, simple_loss=0.3988, pruned_loss=0.1516, over 8495.00 frames. ], batch size: 26, lr: 1.59e-02, grad_scale: 8.0 +2023-02-06 00:13:19,663 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 00:13:31,617 INFO [train.py:935] (1/4) Epoch 5, validation: loss=0.2309, simple_loss=0.3254, pruned_loss=0.06822, over 944034.00 frames. +2023-02-06 00:13:31,618 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-06 00:13:46,430 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 00:13:46,603 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32355.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:14:06,999 INFO [train.py:901] (1/4) Epoch 5, batch 50, loss[loss=0.2828, simple_loss=0.3269, pruned_loss=0.1194, over 8081.00 frames. ], tot_loss[loss=0.3029, simple_loss=0.3616, pruned_loss=0.1221, over 369140.90 frames. ], batch size: 21, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:14:14,087 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:14:22,032 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 00:14:22,903 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 00:14:36,523 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.029e+02 3.148e+02 3.721e+02 4.839e+02 1.477e+03, threshold=7.442e+02, percent-clipped=1.0 +2023-02-06 00:14:38,052 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32427.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:14:41,771 INFO [train.py:901] (1/4) Epoch 5, batch 100, loss[loss=0.2979, simple_loss=0.3532, pruned_loss=0.1213, over 8517.00 frames. ], tot_loss[loss=0.3013, simple_loss=0.3595, pruned_loss=0.1215, over 646985.18 frames. ], batch size: 39, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:14:44,567 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32437.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:14:45,029 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 00:15:01,741 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 00:15:02,111 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32462.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:15:15,803 INFO [train.py:901] (1/4) Epoch 5, batch 150, loss[loss=0.2888, simple_loss=0.3488, pruned_loss=0.1144, over 8026.00 frames. ], tot_loss[loss=0.296, simple_loss=0.3546, pruned_loss=0.1187, over 856675.37 frames. ], batch size: 22, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:15:43,041 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32521.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:15:45,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.082e+02 3.007e+02 3.818e+02 4.644e+02 8.323e+02, threshold=7.636e+02, percent-clipped=1.0 +2023-02-06 00:15:50,802 INFO [train.py:901] (1/4) Epoch 5, batch 200, loss[loss=0.34, simple_loss=0.3856, pruned_loss=0.1472, over 8790.00 frames. ], tot_loss[loss=0.2975, simple_loss=0.3559, pruned_loss=0.1196, over 1020911.69 frames. ], batch size: 49, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:15:59,824 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32546.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:16:06,415 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32556.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:16:08,360 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5507, 2.0628, 2.1666, 1.0234, 2.1385, 1.5145, 0.5377, 1.8266], + device='cuda:1'), covar=tensor([0.0192, 0.0084, 0.0065, 0.0185, 0.0112, 0.0293, 0.0286, 0.0076], + device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0205, 0.0165, 0.0252, 0.0200, 0.0329, 0.0268, 0.0233], + device='cuda:1'), out_proj_covar=tensor([1.1070e-04, 7.7295e-05, 6.0324e-05, 9.2430e-05, 7.6535e-05, 1.3368e-04, + 1.0327e-04, 8.7060e-05], device='cuda:1') +2023-02-06 00:16:23,703 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32581.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:16:24,859 INFO [train.py:901] (1/4) Epoch 5, batch 250, loss[loss=0.269, simple_loss=0.3398, pruned_loss=0.09907, over 8455.00 frames. ], tot_loss[loss=0.2981, simple_loss=0.3563, pruned_loss=0.1199, over 1157405.19 frames. ], batch size: 25, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:16:36,169 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 00:16:45,164 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32611.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:16:46,311 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 00:16:54,483 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 3.241e+02 4.131e+02 4.869e+02 1.219e+03, threshold=8.263e+02, percent-clipped=9.0 +2023-02-06 00:17:00,684 INFO [train.py:901] (1/4) Epoch 5, batch 300, loss[loss=0.2864, simple_loss=0.3271, pruned_loss=0.1229, over 7519.00 frames. ], tot_loss[loss=0.2982, simple_loss=0.3569, pruned_loss=0.1197, over 1259890.21 frames. ], batch size: 18, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:17:03,009 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:17:10,894 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32648.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:17:14,888 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1250, 1.1971, 2.2436, 0.8749, 2.1809, 2.4914, 2.3266, 2.0872], + device='cuda:1'), covar=tensor([0.0994, 0.1095, 0.0492, 0.1938, 0.0518, 0.0328, 0.0498, 0.0745], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0265, 0.0222, 0.0261, 0.0224, 0.0194, 0.0212, 0.0270], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:17:27,580 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32673.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:17:34,188 INFO [train.py:901] (1/4) Epoch 5, batch 350, loss[loss=0.3028, simple_loss=0.3762, pruned_loss=0.1147, over 8440.00 frames. ], tot_loss[loss=0.2982, simple_loss=0.3575, pruned_loss=0.1195, over 1342896.20 frames. ], batch size: 27, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:17:34,409 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32683.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:17:48,303 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3016, 1.1658, 4.4733, 1.5546, 3.7770, 3.7159, 4.0133, 3.9196], + device='cuda:1'), covar=tensor([0.0355, 0.3755, 0.0298, 0.2402, 0.1027, 0.0618, 0.0370, 0.0464], + device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0472, 0.0384, 0.0390, 0.0455, 0.0378, 0.0376, 0.0415], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 00:17:51,715 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32708.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:18:04,034 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.794e+02 3.189e+02 4.031e+02 4.810e+02 8.158e+02, threshold=8.062e+02, percent-clipped=0.0 +2023-02-06 00:18:09,328 INFO [train.py:901] (1/4) Epoch 5, batch 400, loss[loss=0.326, simple_loss=0.3848, pruned_loss=0.1336, over 8344.00 frames. ], tot_loss[loss=0.297, simple_loss=0.3558, pruned_loss=0.1191, over 1396882.35 frames. ], batch size: 26, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:18:43,765 INFO [train.py:901] (1/4) Epoch 5, batch 450, loss[loss=0.3803, simple_loss=0.4014, pruned_loss=0.1796, over 6736.00 frames. ], tot_loss[loss=0.2983, simple_loss=0.3569, pruned_loss=0.1199, over 1445837.69 frames. ], batch size: 71, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:19:02,189 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 00:19:09,314 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0990, 3.0829, 2.8135, 1.5590, 2.7784, 2.7568, 2.9210, 2.5739], + device='cuda:1'), covar=tensor([0.1270, 0.0828, 0.1116, 0.4813, 0.0955, 0.1129, 0.1375, 0.0944], + device='cuda:1'), in_proj_covar=tensor([0.0379, 0.0279, 0.0310, 0.0409, 0.0302, 0.0266, 0.0295, 0.0237], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-06 00:19:12,445 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.263e+02 3.122e+02 4.068e+02 4.898e+02 9.897e+02, threshold=8.137e+02, percent-clipped=5.0 +2023-02-06 00:19:17,681 INFO [train.py:901] (1/4) Epoch 5, batch 500, loss[loss=0.3267, simple_loss=0.3801, pruned_loss=0.1366, over 8238.00 frames. ], tot_loss[loss=0.2987, simple_loss=0.3572, pruned_loss=0.1202, over 1485030.72 frames. ], batch size: 22, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:19:52,898 INFO [train.py:901] (1/4) Epoch 5, batch 550, loss[loss=0.2815, simple_loss=0.3502, pruned_loss=0.1064, over 8191.00 frames. ], tot_loss[loss=0.2974, simple_loss=0.3562, pruned_loss=0.1193, over 1517542.60 frames. ], batch size: 23, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:19:58,693 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4097, 1.9457, 3.4878, 1.0973, 2.4191, 1.7810, 1.5436, 2.0578], + device='cuda:1'), covar=tensor([0.1643, 0.1775, 0.0615, 0.3033, 0.1416, 0.2247, 0.1452, 0.2158], + device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0437, 0.0515, 0.0522, 0.0570, 0.0500, 0.0444, 0.0583], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:20:13,547 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.59 vs. limit=5.0 +2023-02-06 00:20:21,235 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.184e+02 3.133e+02 3.697e+02 5.126e+02 1.321e+03, threshold=7.393e+02, percent-clipped=4.0 +2023-02-06 00:20:26,722 INFO [train.py:901] (1/4) Epoch 5, batch 600, loss[loss=0.2671, simple_loss=0.3181, pruned_loss=0.1081, over 7541.00 frames. ], tot_loss[loss=0.2965, simple_loss=0.3551, pruned_loss=0.119, over 1536540.03 frames. ], batch size: 18, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:20:47,716 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7038, 2.4287, 4.7076, 1.2032, 2.6855, 2.1494, 1.7047, 2.3579], + device='cuda:1'), covar=tensor([0.1321, 0.1526, 0.0510, 0.2830, 0.1297, 0.1979, 0.1270, 0.2136], + device='cuda:1'), in_proj_covar=tensor([0.0464, 0.0440, 0.0518, 0.0527, 0.0571, 0.0502, 0.0445, 0.0587], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:20:50,110 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 00:21:02,150 INFO [train.py:901] (1/4) Epoch 5, batch 650, loss[loss=0.394, simple_loss=0.4216, pruned_loss=0.1831, over 8322.00 frames. ], tot_loss[loss=0.2973, simple_loss=0.356, pruned_loss=0.1193, over 1555704.09 frames. ], batch size: 26, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:21:02,980 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32984.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:21:09,633 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1640, 3.1818, 2.8896, 1.5745, 2.8712, 2.8101, 2.9366, 2.6844], + device='cuda:1'), covar=tensor([0.1280, 0.0806, 0.1168, 0.4474, 0.0951, 0.1072, 0.1422, 0.1056], + device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0277, 0.0312, 0.0405, 0.0300, 0.0264, 0.0293, 0.0236], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-06 00:21:30,780 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 3.090e+02 3.854e+02 5.024e+02 8.355e+02, threshold=7.708e+02, percent-clipped=4.0 +2023-02-06 00:21:36,137 INFO [train.py:901] (1/4) Epoch 5, batch 700, loss[loss=0.2614, simple_loss=0.3347, pruned_loss=0.09399, over 8244.00 frames. ], tot_loss[loss=0.2967, simple_loss=0.3552, pruned_loss=0.1191, over 1567967.15 frames. ], batch size: 24, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:22:11,098 INFO [train.py:901] (1/4) Epoch 5, batch 750, loss[loss=0.295, simple_loss=0.3652, pruned_loss=0.1124, over 8326.00 frames. ], tot_loss[loss=0.2944, simple_loss=0.3534, pruned_loss=0.1176, over 1575648.83 frames. ], batch size: 25, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:22:13,822 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.3168, 1.8314, 1.7718, 1.6547, 1.6385, 1.7666, 2.4741, 2.2880], + device='cuda:1'), covar=tensor([0.0520, 0.1390, 0.2025, 0.1493, 0.0708, 0.1754, 0.0749, 0.0541], + device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0187, 0.0225, 0.0187, 0.0141, 0.0194, 0.0151, 0.0155], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:22:14,421 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:22:36,868 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 00:22:38,059 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.66 vs. limit=5.0 +2023-02-06 00:22:40,967 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.060e+02 3.306e+02 4.079e+02 5.042e+02 1.499e+03, threshold=8.159e+02, percent-clipped=7.0 +2023-02-06 00:22:45,501 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 00:22:46,150 INFO [train.py:901] (1/4) Epoch 5, batch 800, loss[loss=0.268, simple_loss=0.3454, pruned_loss=0.09531, over 8029.00 frames. ], tot_loss[loss=0.2954, simple_loss=0.3542, pruned_loss=0.1183, over 1585007.91 frames. ], batch size: 22, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:23:13,727 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33173.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:23:19,949 INFO [train.py:901] (1/4) Epoch 5, batch 850, loss[loss=0.2811, simple_loss=0.3503, pruned_loss=0.106, over 8335.00 frames. ], tot_loss[loss=0.2956, simple_loss=0.3546, pruned_loss=0.1183, over 1592385.61 frames. ], batch size: 26, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:23:49,971 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.888e+02 3.855e+02 5.468e+02 1.103e+03, threshold=7.709e+02, percent-clipped=2.0 +2023-02-06 00:23:56,031 INFO [train.py:901] (1/4) Epoch 5, batch 900, loss[loss=0.3705, simple_loss=0.4112, pruned_loss=0.1649, over 8329.00 frames. ], tot_loss[loss=0.2953, simple_loss=0.3541, pruned_loss=0.1183, over 1598625.99 frames. ], batch size: 26, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:23:59,531 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7728, 3.7781, 3.3946, 1.8682, 3.4102, 3.2558, 3.4845, 2.8892], + device='cuda:1'), covar=tensor([0.1085, 0.0752, 0.1155, 0.4497, 0.0763, 0.0975, 0.1555, 0.0913], + device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0274, 0.0304, 0.0392, 0.0294, 0.0260, 0.0284, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-06 00:24:29,307 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4696, 1.8642, 3.2032, 1.1803, 2.1666, 1.8285, 1.5188, 1.7697], + device='cuda:1'), covar=tensor([0.1447, 0.1709, 0.0594, 0.2888, 0.1291, 0.2144, 0.1360, 0.2016], + device='cuda:1'), in_proj_covar=tensor([0.0470, 0.0444, 0.0516, 0.0525, 0.0571, 0.0508, 0.0442, 0.0585], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:24:29,755 INFO [train.py:901] (1/4) Epoch 5, batch 950, loss[loss=0.2831, simple_loss=0.3463, pruned_loss=0.11, over 8243.00 frames. ], tot_loss[loss=0.2942, simple_loss=0.3537, pruned_loss=0.1173, over 1601199.00 frames. ], batch size: 24, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:24:54,213 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3973, 1.2924, 4.5895, 1.6290, 3.8235, 3.6874, 3.9456, 3.8781], + device='cuda:1'), covar=tensor([0.0434, 0.4082, 0.0338, 0.2893, 0.1199, 0.0666, 0.0512, 0.0568], + device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0480, 0.0383, 0.0395, 0.0471, 0.0387, 0.0382, 0.0428], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 00:25:01,016 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.004e+02 3.759e+02 4.642e+02 8.675e+02, threshold=7.519e+02, percent-clipped=2.0 +2023-02-06 00:25:03,075 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:25:05,105 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 00:25:06,447 INFO [train.py:901] (1/4) Epoch 5, batch 1000, loss[loss=0.2791, simple_loss=0.3306, pruned_loss=0.1139, over 7926.00 frames. ], tot_loss[loss=0.2936, simple_loss=0.3538, pruned_loss=0.1167, over 1608229.85 frames. ], batch size: 20, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:25:40,440 INFO [train.py:901] (1/4) Epoch 5, batch 1050, loss[loss=0.2948, simple_loss=0.3661, pruned_loss=0.1118, over 8283.00 frames. ], tot_loss[loss=0.2943, simple_loss=0.3544, pruned_loss=0.1171, over 1614569.69 frames. ], batch size: 23, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:25:40,444 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 00:25:42,725 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8231, 1.2471, 3.4092, 1.1886, 2.1760, 3.7497, 3.7102, 3.3431], + device='cuda:1'), covar=tensor([0.1096, 0.1549, 0.0383, 0.2056, 0.0821, 0.0253, 0.0331, 0.0505], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0264, 0.0220, 0.0257, 0.0222, 0.0197, 0.0216, 0.0267], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:25:52,111 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 00:26:08,785 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 3.252e+02 3.786e+02 4.850e+02 9.380e+02, threshold=7.572e+02, percent-clipped=3.0 +2023-02-06 00:26:13,583 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33431.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:26:14,854 INFO [train.py:901] (1/4) Epoch 5, batch 1100, loss[loss=0.2241, simple_loss=0.298, pruned_loss=0.07516, over 8245.00 frames. ], tot_loss[loss=0.2923, simple_loss=0.3527, pruned_loss=0.116, over 1607908.37 frames. ], batch size: 22, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:26:15,001 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9592, 1.7224, 5.9192, 2.3222, 5.2948, 5.1545, 5.5756, 5.4742], + device='cuda:1'), covar=tensor([0.0287, 0.3220, 0.0191, 0.2007, 0.0920, 0.0451, 0.0284, 0.0373], + device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0471, 0.0382, 0.0390, 0.0461, 0.0377, 0.0374, 0.0420], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 00:26:23,005 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33443.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:26:50,043 INFO [train.py:901] (1/4) Epoch 5, batch 1150, loss[loss=0.3062, simple_loss=0.3733, pruned_loss=0.1195, over 8313.00 frames. ], tot_loss[loss=0.2917, simple_loss=0.3521, pruned_loss=0.1157, over 1608400.00 frames. ], batch size: 48, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:26:54,929 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:27:02,048 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 00:27:13,082 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33517.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:27:18,259 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.110e+02 3.101e+02 4.052e+02 5.357e+02 1.331e+03, threshold=8.105e+02, percent-clipped=11.0 +2023-02-06 00:27:23,593 INFO [train.py:901] (1/4) Epoch 5, batch 1200, loss[loss=0.2631, simple_loss=0.3374, pruned_loss=0.09434, over 8512.00 frames. ], tot_loss[loss=0.2936, simple_loss=0.354, pruned_loss=0.1167, over 1612795.43 frames. ], batch size: 28, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:27:32,554 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33546.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:27:55,625 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3117, 1.3853, 1.5331, 1.0877, 1.4603, 1.3236, 1.6984, 1.8172], + device='cuda:1'), covar=tensor([0.0573, 0.1296, 0.1845, 0.1551, 0.0651, 0.1719, 0.0820, 0.0553], + device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0187, 0.0227, 0.0188, 0.0141, 0.0197, 0.0154, 0.0159], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:28:00,122 INFO [train.py:901] (1/4) Epoch 5, batch 1250, loss[loss=0.2643, simple_loss=0.3303, pruned_loss=0.09913, over 8080.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.3552, pruned_loss=0.1179, over 1614007.87 frames. ], batch size: 21, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:28:08,903 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 00:28:29,020 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 3.057e+02 3.737e+02 5.343e+02 1.068e+03, threshold=7.474e+02, percent-clipped=1.0 +2023-02-06 00:28:34,049 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33632.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:28:34,552 INFO [train.py:901] (1/4) Epoch 5, batch 1300, loss[loss=0.3183, simple_loss=0.3762, pruned_loss=0.1302, over 8186.00 frames. ], tot_loss[loss=0.2963, simple_loss=0.3559, pruned_loss=0.1184, over 1619026.40 frames. ], batch size: 23, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:28:38,322 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.79 vs. limit=5.0 +2023-02-06 00:29:10,152 INFO [train.py:901] (1/4) Epoch 5, batch 1350, loss[loss=0.2633, simple_loss=0.3401, pruned_loss=0.09324, over 8501.00 frames. ], tot_loss[loss=0.2962, simple_loss=0.356, pruned_loss=0.1182, over 1619896.78 frames. ], batch size: 28, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:29:18,821 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33695.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:29:21,558 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33699.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:29:38,303 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33724.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:29:39,349 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.845e+02 3.141e+02 3.942e+02 4.566e+02 9.800e+02, threshold=7.885e+02, percent-clipped=1.0 +2023-02-06 00:29:44,191 INFO [train.py:901] (1/4) Epoch 5, batch 1400, loss[loss=0.2458, simple_loss=0.3143, pruned_loss=0.08865, over 7809.00 frames. ], tot_loss[loss=0.2936, simple_loss=0.3541, pruned_loss=0.1166, over 1620039.49 frames. ], batch size: 20, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:30:18,165 INFO [train.py:901] (1/4) Epoch 5, batch 1450, loss[loss=0.3554, simple_loss=0.4084, pruned_loss=0.1512, over 8590.00 frames. ], tot_loss[loss=0.2948, simple_loss=0.3548, pruned_loss=0.1174, over 1615930.54 frames. ], batch size: 31, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:30:32,271 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 00:30:32,510 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:30:49,117 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.073e+02 3.068e+02 3.705e+02 5.190e+02 1.303e+03, threshold=7.410e+02, percent-clipped=4.0 +2023-02-06 00:30:50,028 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33827.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:30:54,026 INFO [train.py:901] (1/4) Epoch 5, batch 1500, loss[loss=0.2578, simple_loss=0.3253, pruned_loss=0.09516, over 7813.00 frames. ], tot_loss[loss=0.2933, simple_loss=0.3537, pruned_loss=0.1165, over 1617027.08 frames. ], batch size: 20, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:30:54,780 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33834.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:31:27,613 INFO [train.py:901] (1/4) Epoch 5, batch 1550, loss[loss=0.2268, simple_loss=0.3031, pruned_loss=0.07527, over 7822.00 frames. ], tot_loss[loss=0.2936, simple_loss=0.3541, pruned_loss=0.1165, over 1621154.75 frames. ], batch size: 19, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:31:31,089 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33888.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:31:34,152 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-02-06 00:31:48,411 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33913.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:31:51,173 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-02-06 00:31:58,130 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 3.419e+02 4.027e+02 4.998e+02 8.696e+02, threshold=8.054e+02, percent-clipped=2.0 +2023-02-06 00:32:03,155 INFO [train.py:901] (1/4) Epoch 5, batch 1600, loss[loss=0.3463, simple_loss=0.3927, pruned_loss=0.15, over 8465.00 frames. ], tot_loss[loss=0.2962, simple_loss=0.356, pruned_loss=0.1181, over 1618882.56 frames. ], batch size: 39, lr: 1.55e-02, grad_scale: 8.0 +2023-02-06 00:32:14,804 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33949.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:32:20,119 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33957.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:32:28,683 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7037, 1.8867, 1.5207, 2.3591, 1.2138, 1.4249, 1.5907, 1.9017], + device='cuda:1'), covar=tensor([0.1035, 0.1351, 0.1705, 0.0605, 0.1761, 0.2164, 0.1454, 0.1110], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0274, 0.0292, 0.0226, 0.0260, 0.0292, 0.0289, 0.0269], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:32:37,058 INFO [train.py:901] (1/4) Epoch 5, batch 1650, loss[loss=0.267, simple_loss=0.3459, pruned_loss=0.09399, over 8101.00 frames. ], tot_loss[loss=0.2941, simple_loss=0.3544, pruned_loss=0.1169, over 1620012.42 frames. ], batch size: 23, lr: 1.55e-02, grad_scale: 8.0 +2023-02-06 00:33:07,303 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.007e+02 2.926e+02 3.722e+02 4.611e+02 9.053e+02, threshold=7.444e+02, percent-clipped=4.0 +2023-02-06 00:33:11,841 INFO [train.py:901] (1/4) Epoch 5, batch 1700, loss[loss=0.2746, simple_loss=0.3438, pruned_loss=0.1027, over 8248.00 frames. ], tot_loss[loss=0.2947, simple_loss=0.3549, pruned_loss=0.1173, over 1620601.75 frames. ], batch size: 24, lr: 1.55e-02, grad_scale: 8.0 +2023-02-06 00:33:17,266 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34039.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:33:33,625 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:33:42,178 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 00:33:47,846 INFO [train.py:901] (1/4) Epoch 5, batch 1750, loss[loss=0.2544, simple_loss=0.3177, pruned_loss=0.0956, over 7260.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.3544, pruned_loss=0.1174, over 1614839.84 frames. ], batch size: 16, lr: 1.55e-02, grad_scale: 8.0 +2023-02-06 00:34:16,525 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.177e+02 3.118e+02 3.687e+02 4.787e+02 9.448e+02, threshold=7.373e+02, percent-clipped=7.0 +2023-02-06 00:34:21,854 INFO [train.py:901] (1/4) Epoch 5, batch 1800, loss[loss=0.3276, simple_loss=0.3777, pruned_loss=0.1388, over 8598.00 frames. ], tot_loss[loss=0.2975, simple_loss=0.3565, pruned_loss=0.1192, over 1619556.40 frames. ], batch size: 31, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:34:36,712 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34154.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:34:41,453 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 00:34:43,251 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34163.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:34:57,299 INFO [train.py:901] (1/4) Epoch 5, batch 1850, loss[loss=0.3059, simple_loss=0.3616, pruned_loss=0.1251, over 8561.00 frames. ], tot_loss[loss=0.297, simple_loss=0.3562, pruned_loss=0.1188, over 1623324.83 frames. ], batch size: 31, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:35:12,358 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34205.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:35:26,216 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 3.489e+02 4.150e+02 5.670e+02 1.027e+03, threshold=8.299e+02, percent-clipped=7.0 +2023-02-06 00:35:29,077 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34230.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:35:30,902 INFO [train.py:901] (1/4) Epoch 5, batch 1900, loss[loss=0.2682, simple_loss=0.3366, pruned_loss=0.09989, over 8282.00 frames. ], tot_loss[loss=0.2967, simple_loss=0.3554, pruned_loss=0.119, over 1621250.06 frames. ], batch size: 23, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:35:45,547 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7029, 1.6413, 3.0115, 1.2266, 2.1853, 3.3398, 3.2460, 2.8308], + device='cuda:1'), covar=tensor([0.0966, 0.1213, 0.0395, 0.1894, 0.0656, 0.0238, 0.0352, 0.0609], + device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0260, 0.0215, 0.0255, 0.0215, 0.0192, 0.0217, 0.0264], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:35:56,409 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.33 vs. limit=5.0 +2023-02-06 00:36:05,943 INFO [train.py:901] (1/4) Epoch 5, batch 1950, loss[loss=0.2733, simple_loss=0.3454, pruned_loss=0.1006, over 8515.00 frames. ], tot_loss[loss=0.2973, simple_loss=0.356, pruned_loss=0.1194, over 1624635.26 frames. ], batch size: 28, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:36:09,867 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 00:36:18,716 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34301.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:36:23,384 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 00:36:35,388 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.917e+02 2.967e+02 3.945e+02 4.927e+02 1.257e+03, threshold=7.890e+02, percent-clipped=2.0 +2023-02-06 00:36:40,168 INFO [train.py:901] (1/4) Epoch 5, batch 2000, loss[loss=0.3325, simple_loss=0.3937, pruned_loss=0.1356, over 8767.00 frames. ], tot_loss[loss=0.2965, simple_loss=0.3558, pruned_loss=0.1187, over 1623184.41 frames. ], batch size: 30, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:36:42,291 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 00:37:14,372 INFO [train.py:901] (1/4) Epoch 5, batch 2050, loss[loss=0.2799, simple_loss=0.3339, pruned_loss=0.113, over 8090.00 frames. ], tot_loss[loss=0.2963, simple_loss=0.3555, pruned_loss=0.1185, over 1621762.11 frames. ], batch size: 21, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:37:26,355 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.9849, 3.9856, 3.6525, 1.9178, 3.5938, 3.6160, 3.7167, 3.2919], + device='cuda:1'), covar=tensor([0.1130, 0.0618, 0.0993, 0.4972, 0.0964, 0.0931, 0.1177, 0.0971], + device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0280, 0.0300, 0.0392, 0.0302, 0.0265, 0.0290, 0.0238], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-06 00:37:31,119 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34406.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:37:33,927 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34410.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:37:38,680 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34416.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:37:45,186 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 3.245e+02 4.066e+02 4.898e+02 1.293e+03, threshold=8.132e+02, percent-clipped=4.0 +2023-02-06 00:37:49,849 INFO [train.py:901] (1/4) Epoch 5, batch 2100, loss[loss=0.3371, simple_loss=0.3832, pruned_loss=0.1455, over 8687.00 frames. ], tot_loss[loss=0.2956, simple_loss=0.3549, pruned_loss=0.1182, over 1618399.23 frames. ], batch size: 39, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:37:50,883 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 +2023-02-06 00:37:51,417 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:37:52,359 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.45 vs. limit=5.0 +2023-02-06 00:38:08,191 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3787, 1.5556, 1.5293, 0.7694, 1.5892, 1.2673, 0.6799, 1.5253], + device='cuda:1'), covar=tensor([0.0158, 0.0096, 0.0079, 0.0150, 0.0099, 0.0232, 0.0226, 0.0071], + device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0213, 0.0175, 0.0259, 0.0207, 0.0345, 0.0273, 0.0246], + device='cuda:1'), out_proj_covar=tensor([1.1004e-04, 7.8018e-05, 6.2492e-05, 9.2994e-05, 7.7861e-05, 1.3715e-04, + 1.0192e-04, 9.0270e-05], device='cuda:1') +2023-02-06 00:38:15,732 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.15 vs. limit=5.0 +2023-02-06 00:38:23,173 INFO [train.py:901] (1/4) Epoch 5, batch 2150, loss[loss=0.2482, simple_loss=0.3119, pruned_loss=0.09228, over 7419.00 frames. ], tot_loss[loss=0.2945, simple_loss=0.3534, pruned_loss=0.1178, over 1611811.85 frames. ], batch size: 17, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:38:39,338 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9322, 1.4757, 3.4344, 1.3470, 2.1555, 3.7723, 3.5954, 3.2145], + device='cuda:1'), covar=tensor([0.1018, 0.1331, 0.0308, 0.1910, 0.0733, 0.0227, 0.0379, 0.0591], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0262, 0.0220, 0.0259, 0.0218, 0.0196, 0.0226, 0.0273], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 00:38:39,926 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:38:50,623 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34521.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:38:53,807 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.929e+02 3.753e+02 4.663e+02 1.529e+03, threshold=7.506e+02, percent-clipped=2.0 +2023-02-06 00:38:59,147 INFO [train.py:901] (1/4) Epoch 5, batch 2200, loss[loss=0.4335, simple_loss=0.4492, pruned_loss=0.2089, over 7208.00 frames. ], tot_loss[loss=0.2947, simple_loss=0.3535, pruned_loss=0.118, over 1611979.76 frames. ], batch size: 72, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:39:02,751 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4792, 2.0839, 4.2853, 1.0897, 2.8343, 1.9240, 1.6606, 2.3268], + device='cuda:1'), covar=tensor([0.1517, 0.1818, 0.0790, 0.3218, 0.1372, 0.2492, 0.1349, 0.2438], + device='cuda:1'), in_proj_covar=tensor([0.0464, 0.0441, 0.0526, 0.0525, 0.0568, 0.0512, 0.0441, 0.0575], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:39:32,427 INFO [train.py:901] (1/4) Epoch 5, batch 2250, loss[loss=0.2735, simple_loss=0.3031, pruned_loss=0.122, over 7420.00 frames. ], tot_loss[loss=0.2966, simple_loss=0.3549, pruned_loss=0.1191, over 1610246.77 frames. ], batch size: 17, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:39:52,221 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34611.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:39:59,498 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:40:02,635 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 3.688e+02 4.883e+02 6.349e+02 4.437e+03, threshold=9.766e+02, percent-clipped=16.0 +2023-02-06 00:40:07,918 INFO [train.py:901] (1/4) Epoch 5, batch 2300, loss[loss=0.3042, simple_loss=0.368, pruned_loss=0.1202, over 8442.00 frames. ], tot_loss[loss=0.2984, simple_loss=0.3562, pruned_loss=0.1203, over 1610098.75 frames. ], batch size: 27, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:40:12,724 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34640.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:40:16,065 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:40:34,989 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:40:42,251 INFO [train.py:901] (1/4) Epoch 5, batch 2350, loss[loss=0.2625, simple_loss=0.3355, pruned_loss=0.0947, over 8102.00 frames. ], tot_loss[loss=0.2978, simple_loss=0.356, pruned_loss=0.1198, over 1608474.60 frames. ], batch size: 23, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:40:51,691 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34697.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:41:11,432 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.800e+02 3.189e+02 4.018e+02 4.942e+02 1.178e+03, threshold=8.036e+02, percent-clipped=1.0 +2023-02-06 00:41:12,930 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34728.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:41:16,111 INFO [train.py:901] (1/4) Epoch 5, batch 2400, loss[loss=0.2724, simple_loss=0.351, pruned_loss=0.09688, over 8473.00 frames. ], tot_loss[loss=0.2961, simple_loss=0.3545, pruned_loss=0.1188, over 1608671.29 frames. ], batch size: 25, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:41:47,417 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34777.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:41:51,224 INFO [train.py:901] (1/4) Epoch 5, batch 2450, loss[loss=0.2853, simple_loss=0.3614, pruned_loss=0.1046, over 8448.00 frames. ], tot_loss[loss=0.2957, simple_loss=0.3544, pruned_loss=0.1185, over 1611823.63 frames. ], batch size: 25, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:42:04,202 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:42:19,460 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 2.991e+02 3.791e+02 4.954e+02 1.109e+03, threshold=7.583e+02, percent-clipped=3.0 +2023-02-06 00:42:24,154 INFO [train.py:901] (1/4) Epoch 5, batch 2500, loss[loss=0.2817, simple_loss=0.3514, pruned_loss=0.106, over 8462.00 frames. ], tot_loss[loss=0.2943, simple_loss=0.3534, pruned_loss=0.1176, over 1614511.49 frames. ], batch size: 25, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:42:29,096 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4008, 1.8644, 1.9131, 0.8385, 1.9022, 1.4808, 0.4299, 1.5877], + device='cuda:1'), covar=tensor([0.0216, 0.0109, 0.0098, 0.0178, 0.0147, 0.0343, 0.0290, 0.0093], + device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0215, 0.0178, 0.0258, 0.0207, 0.0347, 0.0275, 0.0247], + device='cuda:1'), out_proj_covar=tensor([1.1089e-04, 7.8472e-05, 6.3019e-05, 9.2484e-05, 7.7289e-05, 1.3744e-04, + 1.0226e-04, 9.0032e-05], device='cuda:1') +2023-02-06 00:42:53,616 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 00:42:55,973 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34878.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:42:59,771 INFO [train.py:901] (1/4) Epoch 5, batch 2550, loss[loss=0.2713, simple_loss=0.3336, pruned_loss=0.1045, over 7926.00 frames. ], tot_loss[loss=0.2961, simple_loss=0.3547, pruned_loss=0.1187, over 1614534.83 frames. ], batch size: 20, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:43:13,493 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34903.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:43:29,143 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.077e+02 2.947e+02 3.618e+02 4.736e+02 1.253e+03, threshold=7.237e+02, percent-clipped=4.0 +2023-02-06 00:43:31,377 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0284, 1.0069, 4.2388, 1.5657, 3.6107, 3.5502, 3.7833, 3.7023], + device='cuda:1'), covar=tensor([0.0408, 0.3802, 0.0302, 0.2303, 0.0951, 0.0549, 0.0419, 0.0527], + device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0492, 0.0403, 0.0413, 0.0482, 0.0397, 0.0400, 0.0444], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 00:43:33,924 INFO [train.py:901] (1/4) Epoch 5, batch 2600, loss[loss=0.3353, simple_loss=0.3926, pruned_loss=0.139, over 8110.00 frames. ], tot_loss[loss=0.2948, simple_loss=0.3537, pruned_loss=0.118, over 1612955.21 frames. ], batch size: 23, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:43:40,976 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2969, 1.7094, 1.6201, 0.5277, 1.7263, 1.2304, 0.2979, 1.5624], + device='cuda:1'), covar=tensor([0.0167, 0.0091, 0.0077, 0.0166, 0.0086, 0.0307, 0.0237, 0.0065], + device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0217, 0.0179, 0.0261, 0.0206, 0.0350, 0.0275, 0.0249], + device='cuda:1'), out_proj_covar=tensor([1.1228e-04, 7.9056e-05, 6.3550e-05, 9.3850e-05, 7.7019e-05, 1.3833e-04, + 1.0169e-04, 9.0907e-05], device='cuda:1') +2023-02-06 00:43:49,011 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34955.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:44:09,533 INFO [train.py:901] (1/4) Epoch 5, batch 2650, loss[loss=0.2718, simple_loss=0.3236, pruned_loss=0.11, over 7711.00 frames. ], tot_loss[loss=0.2935, simple_loss=0.3529, pruned_loss=0.1171, over 1613863.05 frames. ], batch size: 18, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:44:10,276 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34984.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:44:13,022 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34988.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:44:21,474 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0638, 2.2822, 4.0078, 3.1061, 3.0528, 2.3661, 1.6032, 1.7686], + device='cuda:1'), covar=tensor([0.1603, 0.2285, 0.0428, 0.0897, 0.0933, 0.0945, 0.1137, 0.2100], + device='cuda:1'), in_proj_covar=tensor([0.0743, 0.0668, 0.0571, 0.0650, 0.0751, 0.0626, 0.0602, 0.0625], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 00:44:39,310 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.917e+02 2.922e+02 3.827e+02 4.980e+02 8.274e+02, threshold=7.654e+02, percent-clipped=5.0 +2023-02-06 00:44:43,771 INFO [train.py:901] (1/4) Epoch 5, batch 2700, loss[loss=0.3096, simple_loss=0.3683, pruned_loss=0.1254, over 8188.00 frames. ], tot_loss[loss=0.2935, simple_loss=0.3531, pruned_loss=0.1169, over 1610426.10 frames. ], batch size: 23, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:45:09,481 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35070.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:45:10,725 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35072.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:45:18,313 INFO [train.py:901] (1/4) Epoch 5, batch 2750, loss[loss=0.2908, simple_loss=0.3649, pruned_loss=0.1083, over 8289.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3521, pruned_loss=0.1159, over 1607302.21 frames. ], batch size: 23, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:45:24,479 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7366, 1.8534, 2.0757, 1.6069, 1.0742, 2.0555, 0.3494, 1.2981], + device='cuda:1'), covar=tensor([0.3819, 0.1974, 0.0999, 0.3176, 0.6587, 0.0822, 0.5821, 0.2310], + device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0126, 0.0079, 0.0166, 0.0207, 0.0081, 0.0146, 0.0121], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 00:45:29,768 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35099.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:45:32,528 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35103.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:45:37,954 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35110.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:45:48,688 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 3.039e+02 3.659e+02 5.251e+02 1.248e+03, threshold=7.317e+02, percent-clipped=8.0 +2023-02-06 00:45:53,678 INFO [train.py:901] (1/4) Epoch 5, batch 2800, loss[loss=0.225, simple_loss=0.2941, pruned_loss=0.07793, over 7530.00 frames. ], tot_loss[loss=0.292, simple_loss=0.3516, pruned_loss=0.1162, over 1603134.33 frames. ], batch size: 18, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:46:27,525 INFO [train.py:901] (1/4) Epoch 5, batch 2850, loss[loss=0.2513, simple_loss=0.3124, pruned_loss=0.09512, over 7526.00 frames. ], tot_loss[loss=0.291, simple_loss=0.3511, pruned_loss=0.1154, over 1606976.38 frames. ], batch size: 18, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:46:30,533 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35187.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:46:58,361 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.147e+02 2.990e+02 3.598e+02 4.675e+02 1.498e+03, threshold=7.197e+02, percent-clipped=4.0 +2023-02-06 00:47:03,666 INFO [train.py:901] (1/4) Epoch 5, batch 2900, loss[loss=0.3412, simple_loss=0.3828, pruned_loss=0.1498, over 7982.00 frames. ], tot_loss[loss=0.2923, simple_loss=0.3519, pruned_loss=0.1163, over 1609772.38 frames. ], batch size: 21, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:47:36,536 INFO [train.py:901] (1/4) Epoch 5, batch 2950, loss[loss=0.3116, simple_loss=0.376, pruned_loss=0.1236, over 8186.00 frames. ], tot_loss[loss=0.2928, simple_loss=0.3527, pruned_loss=0.1164, over 1609841.00 frames. ], batch size: 23, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:47:41,854 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 00:48:06,831 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 3.185e+02 3.825e+02 4.988e+02 1.295e+03, threshold=7.649e+02, percent-clipped=4.0 +2023-02-06 00:48:07,076 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:48:12,108 INFO [train.py:901] (1/4) Epoch 5, batch 3000, loss[loss=0.2497, simple_loss=0.3261, pruned_loss=0.08663, over 8099.00 frames. ], tot_loss[loss=0.2927, simple_loss=0.3528, pruned_loss=0.1163, over 1614040.89 frames. ], batch size: 23, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:48:12,108 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 00:48:25,515 INFO [train.py:935] (1/4) Epoch 5, validation: loss=0.2228, simple_loss=0.319, pruned_loss=0.0633, over 944034.00 frames. +2023-02-06 00:48:25,516 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-06 00:48:39,251 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35351.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:48:41,188 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0525, 4.0422, 3.6970, 1.7839, 3.6477, 3.5964, 3.7762, 3.0746], + device='cuda:1'), covar=tensor([0.1008, 0.0635, 0.0948, 0.4513, 0.0804, 0.0692, 0.1268, 0.0936], + device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0280, 0.0314, 0.0398, 0.0309, 0.0270, 0.0295, 0.0240], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-06 00:48:41,991 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35355.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:48:43,991 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9724, 1.4640, 1.4057, 1.2986, 1.2602, 1.3826, 1.5149, 1.3355], + device='cuda:1'), covar=tensor([0.0614, 0.1200, 0.1849, 0.1384, 0.0658, 0.1520, 0.0802, 0.0654], + device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0184, 0.0223, 0.0185, 0.0138, 0.0193, 0.0148, 0.0156], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 00:48:44,701 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:48:47,343 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35363.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:48:59,226 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35380.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:49:01,094 INFO [train.py:901] (1/4) Epoch 5, batch 3050, loss[loss=0.273, simple_loss=0.3409, pruned_loss=0.1025, over 8580.00 frames. ], tot_loss[loss=0.294, simple_loss=0.3542, pruned_loss=0.1169, over 1619867.20 frames. ], batch size: 39, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:49:01,964 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35384.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:49:07,185 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:49:29,668 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.238e+02 3.010e+02 3.735e+02 4.816e+02 9.592e+02, threshold=7.471e+02, percent-clipped=3.0 +2023-02-06 00:49:34,197 INFO [train.py:901] (1/4) Epoch 5, batch 3100, loss[loss=0.3128, simple_loss=0.3777, pruned_loss=0.1239, over 8581.00 frames. ], tot_loss[loss=0.2929, simple_loss=0.3537, pruned_loss=0.1161, over 1625476.45 frames. ], batch size: 34, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:49:41,040 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35443.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:49:48,047 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35454.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:49:59,664 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35468.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:50:09,995 INFO [train.py:901] (1/4) Epoch 5, batch 3150, loss[loss=0.3585, simple_loss=0.4148, pruned_loss=0.1511, over 8454.00 frames. ], tot_loss[loss=0.294, simple_loss=0.3542, pruned_loss=0.1169, over 1622465.20 frames. ], batch size: 27, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:50:21,416 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-06 00:50:39,624 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.108e+02 3.249e+02 4.087e+02 5.030e+02 9.472e+02, threshold=8.174e+02, percent-clipped=3.0 +2023-02-06 00:50:44,432 INFO [train.py:901] (1/4) Epoch 5, batch 3200, loss[loss=0.2669, simple_loss=0.326, pruned_loss=0.1039, over 7814.00 frames. ], tot_loss[loss=0.2935, simple_loss=0.3538, pruned_loss=0.1166, over 1623642.81 frames. ], batch size: 19, lr: 1.51e-02, grad_scale: 8.0 +2023-02-06 00:50:54,842 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35548.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:51:09,483 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35569.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:51:20,169 INFO [train.py:901] (1/4) Epoch 5, batch 3250, loss[loss=0.2578, simple_loss=0.3261, pruned_loss=0.0947, over 7520.00 frames. ], tot_loss[loss=0.2938, simple_loss=0.3543, pruned_loss=0.1167, over 1624383.25 frames. ], batch size: 18, lr: 1.51e-02, grad_scale: 8.0 +2023-02-06 00:51:50,106 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5571, 1.8501, 3.6146, 1.1672, 2.4728, 1.8730, 1.6296, 2.1164], + device='cuda:1'), covar=tensor([0.1405, 0.1894, 0.0612, 0.2908, 0.1356, 0.2212, 0.1354, 0.2211], + device='cuda:1'), in_proj_covar=tensor([0.0473, 0.0451, 0.0530, 0.0541, 0.0581, 0.0523, 0.0458, 0.0595], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 00:51:50,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 3.378e+02 4.149e+02 5.121e+02 1.146e+03, threshold=8.298e+02, percent-clipped=3.0 +2023-02-06 00:51:55,293 INFO [train.py:901] (1/4) Epoch 5, batch 3300, loss[loss=0.2354, simple_loss=0.2975, pruned_loss=0.08665, over 7715.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.3543, pruned_loss=0.1175, over 1620650.77 frames. ], batch size: 18, lr: 1.51e-02, grad_scale: 8.0 +2023-02-06 00:52:07,279 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3540, 2.0590, 4.4616, 2.0342, 2.5165, 5.1378, 4.9871, 4.4973], + device='cuda:1'), covar=tensor([0.1095, 0.1252, 0.0244, 0.1744, 0.0740, 0.0189, 0.0258, 0.0479], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0262, 0.0219, 0.0263, 0.0218, 0.0195, 0.0227, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 00:52:21,826 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3886, 2.0246, 3.1473, 2.3531, 2.6542, 2.0122, 1.4362, 1.3620], + device='cuda:1'), covar=tensor([0.1739, 0.1956, 0.0401, 0.1009, 0.0816, 0.0996, 0.1159, 0.1888], + device='cuda:1'), in_proj_covar=tensor([0.0733, 0.0664, 0.0571, 0.0653, 0.0755, 0.0624, 0.0603, 0.0617], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 00:52:30,161 INFO [train.py:901] (1/4) Epoch 5, batch 3350, loss[loss=0.2506, simple_loss=0.3241, pruned_loss=0.08851, over 8198.00 frames. ], tot_loss[loss=0.293, simple_loss=0.3529, pruned_loss=0.1165, over 1617211.03 frames. ], batch size: 23, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:52:47,838 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35707.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:53:01,472 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 3.555e+02 4.125e+02 4.946e+02 1.065e+03, threshold=8.250e+02, percent-clipped=5.0 +2023-02-06 00:53:06,244 INFO [train.py:901] (1/4) Epoch 5, batch 3400, loss[loss=0.307, simple_loss=0.3637, pruned_loss=0.1251, over 8250.00 frames. ], tot_loss[loss=0.2917, simple_loss=0.3516, pruned_loss=0.1159, over 1613860.71 frames. ], batch size: 24, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:53:08,410 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:53:39,726 INFO [train.py:901] (1/4) Epoch 5, batch 3450, loss[loss=0.2542, simple_loss=0.3264, pruned_loss=0.09099, over 7931.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3519, pruned_loss=0.116, over 1612336.75 frames. ], batch size: 20, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:53:43,885 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:54:07,868 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35822.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:54:09,936 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35825.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:54:10,371 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.123e+02 3.051e+02 3.738e+02 4.571e+02 6.690e+02, threshold=7.475e+02, percent-clipped=0.0 +2023-02-06 00:54:12,579 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35829.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:54:15,094 INFO [train.py:901] (1/4) Epoch 5, batch 3500, loss[loss=0.2047, simple_loss=0.2792, pruned_loss=0.06504, over 7278.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3511, pruned_loss=0.1147, over 1612360.81 frames. ], batch size: 16, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:54:27,027 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35850.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:54:27,681 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:54:40,689 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 00:54:48,819 INFO [train.py:901] (1/4) Epoch 5, batch 3550, loss[loss=0.2165, simple_loss=0.291, pruned_loss=0.07104, over 8245.00 frames. ], tot_loss[loss=0.289, simple_loss=0.3501, pruned_loss=0.1139, over 1616108.17 frames. ], batch size: 24, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:54:54,909 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35892.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:55:19,537 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.232e+02 3.318e+02 3.882e+02 4.908e+02 1.221e+03, threshold=7.763e+02, percent-clipped=6.0 +2023-02-06 00:55:23,582 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5874, 1.9581, 2.2723, 0.9731, 2.2974, 1.3927, 0.5856, 1.6970], + device='cuda:1'), covar=tensor([0.0184, 0.0097, 0.0068, 0.0175, 0.0076, 0.0314, 0.0251, 0.0089], + device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0223, 0.0178, 0.0269, 0.0205, 0.0352, 0.0276, 0.0249], + device='cuda:1'), out_proj_covar=tensor([1.1077e-04, 8.0166e-05, 6.2543e-05, 9.5879e-05, 7.4915e-05, 1.3728e-04, + 1.0157e-04, 8.9133e-05], device='cuda:1') +2023-02-06 00:55:23,994 INFO [train.py:901] (1/4) Epoch 5, batch 3600, loss[loss=0.309, simple_loss=0.378, pruned_loss=0.12, over 8250.00 frames. ], tot_loss[loss=0.2882, simple_loss=0.3494, pruned_loss=0.1135, over 1614130.29 frames. ], batch size: 24, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:55:33,166 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-02-06 00:55:34,957 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2569, 1.4678, 1.4310, 1.3075, 1.2924, 1.3642, 1.7329, 1.5378], + device='cuda:1'), covar=tensor([0.0587, 0.1194, 0.1958, 0.1400, 0.0622, 0.1585, 0.0723, 0.0563], + device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0183, 0.0225, 0.0187, 0.0137, 0.0194, 0.0147, 0.0159], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 00:55:57,699 INFO [train.py:901] (1/4) Epoch 5, batch 3650, loss[loss=0.3075, simple_loss=0.3529, pruned_loss=0.131, over 7523.00 frames. ], tot_loss[loss=0.2906, simple_loss=0.3518, pruned_loss=0.1147, over 1618205.56 frames. ], batch size: 18, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:56:15,071 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36007.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:56:19,743 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4876, 4.5536, 3.9329, 1.8867, 4.0019, 3.8969, 4.1854, 3.4555], + device='cuda:1'), covar=tensor([0.0687, 0.0462, 0.0887, 0.4512, 0.0791, 0.0849, 0.0968, 0.0751], + device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0283, 0.0316, 0.0401, 0.0313, 0.0275, 0.0298, 0.0244], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-06 00:56:27,673 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.237e+02 3.345e+02 4.197e+02 5.280e+02 9.599e+02, threshold=8.394e+02, percent-clipped=10.0 +2023-02-06 00:56:32,340 INFO [train.py:901] (1/4) Epoch 5, batch 3700, loss[loss=0.2757, simple_loss=0.3214, pruned_loss=0.115, over 7922.00 frames. ], tot_loss[loss=0.2927, simple_loss=0.353, pruned_loss=0.1162, over 1617175.47 frames. ], batch size: 20, lr: 1.50e-02, grad_scale: 16.0 +2023-02-06 00:56:40,787 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 00:57:04,809 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36078.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:57:07,931 INFO [train.py:901] (1/4) Epoch 5, batch 3750, loss[loss=0.2553, simple_loss=0.3322, pruned_loss=0.08923, over 8300.00 frames. ], tot_loss[loss=0.2921, simple_loss=0.3527, pruned_loss=0.1158, over 1615538.66 frames. ], batch size: 23, lr: 1.50e-02, grad_scale: 16.0 +2023-02-06 00:57:21,381 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36103.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:57:24,136 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:57:28,741 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3245, 1.5107, 1.7168, 1.3417, 1.3725, 1.4923, 1.7793, 1.7735], + device='cuda:1'), covar=tensor([0.0627, 0.1346, 0.1741, 0.1476, 0.0660, 0.1540, 0.0816, 0.0585], + device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0181, 0.0222, 0.0185, 0.0136, 0.0191, 0.0148, 0.0157], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 00:57:37,193 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.699e+02 3.033e+02 3.704e+02 4.599e+02 1.470e+03, threshold=7.408e+02, percent-clipped=9.0 +2023-02-06 00:57:40,782 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36132.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:57:41,243 INFO [train.py:901] (1/4) Epoch 5, batch 3800, loss[loss=0.3348, simple_loss=0.3869, pruned_loss=0.1414, over 8324.00 frames. ], tot_loss[loss=0.2921, simple_loss=0.3523, pruned_loss=0.1159, over 1614416.69 frames. ], batch size: 25, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 00:57:41,310 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36133.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:57:44,350 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 00:58:09,342 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36173.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:58:15,719 INFO [train.py:901] (1/4) Epoch 5, batch 3850, loss[loss=0.3463, simple_loss=0.4032, pruned_loss=0.1447, over 8578.00 frames. ], tot_loss[loss=0.2932, simple_loss=0.3534, pruned_loss=0.1165, over 1616097.98 frames. ], batch size: 31, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 00:58:27,210 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36199.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:58:41,068 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 00:58:45,745 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 3.284e+02 4.097e+02 5.243e+02 1.380e+03, threshold=8.194e+02, percent-clipped=10.0 +2023-02-06 00:58:49,717 INFO [train.py:901] (1/4) Epoch 5, batch 3900, loss[loss=0.306, simple_loss=0.3657, pruned_loss=0.1232, over 8569.00 frames. ], tot_loss[loss=0.2914, simple_loss=0.3519, pruned_loss=0.1155, over 1615539.79 frames. ], batch size: 31, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 00:58:51,248 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3127, 1.3353, 4.4537, 1.7872, 3.6696, 3.7011, 3.9093, 3.8753], + device='cuda:1'), covar=tensor([0.0388, 0.3850, 0.0366, 0.2378, 0.1263, 0.0613, 0.0462, 0.0554], + device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0498, 0.0414, 0.0419, 0.0493, 0.0407, 0.0398, 0.0450], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 00:58:59,605 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36248.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:59:09,648 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36263.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:59:24,827 INFO [train.py:901] (1/4) Epoch 5, batch 3950, loss[loss=0.2661, simple_loss=0.3316, pruned_loss=0.1003, over 7649.00 frames. ], tot_loss[loss=0.2905, simple_loss=0.3508, pruned_loss=0.1151, over 1611480.21 frames. ], batch size: 19, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 00:59:28,384 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:59:28,408 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:59:35,016 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0579, 3.2905, 3.4197, 2.6077, 1.8573, 3.1845, 0.8536, 2.0531], + device='cuda:1'), covar=tensor([0.6981, 0.1531, 0.0676, 0.2104, 0.4766, 0.1355, 0.6653, 0.2561], + device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0129, 0.0082, 0.0171, 0.0212, 0.0085, 0.0149, 0.0126], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 00:59:45,899 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5461, 4.6070, 3.9778, 1.6251, 3.8780, 4.0236, 4.2389, 3.4869], + device='cuda:1'), covar=tensor([0.0748, 0.0453, 0.0916, 0.4898, 0.0754, 0.0624, 0.1135, 0.0823], + device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0282, 0.0323, 0.0406, 0.0315, 0.0273, 0.0304, 0.0247], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 00:59:51,387 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1248, 3.1060, 2.8286, 1.3572, 2.7484, 2.6507, 2.9151, 2.4593], + device='cuda:1'), covar=tensor([0.1217, 0.0833, 0.1306, 0.4450, 0.1067, 0.1166, 0.1482, 0.1031], + device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0283, 0.0321, 0.0406, 0.0315, 0.0274, 0.0303, 0.0247], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-02-06 00:59:54,576 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.959e+02 3.540e+02 4.519e+02 1.633e+03, threshold=7.079e+02, percent-clipped=6.0 +2023-02-06 00:59:58,373 INFO [train.py:901] (1/4) Epoch 5, batch 4000, loss[loss=0.3118, simple_loss=0.3697, pruned_loss=0.127, over 8516.00 frames. ], tot_loss[loss=0.2922, simple_loss=0.3518, pruned_loss=0.1163, over 1609012.59 frames. ], batch size: 28, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 01:00:19,939 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0984, 3.3854, 2.4948, 4.2172, 2.1150, 2.3379, 2.7228, 3.5521], + device='cuda:1'), covar=tensor([0.0849, 0.0901, 0.1437, 0.0244, 0.1554, 0.1935, 0.1483, 0.0734], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0259, 0.0284, 0.0225, 0.0251, 0.0285, 0.0286, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 01:00:22,757 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4997, 1.9761, 3.1769, 2.5816, 2.5406, 2.0283, 1.4460, 1.1912], + device='cuda:1'), covar=tensor([0.1597, 0.2072, 0.0385, 0.0966, 0.0911, 0.0920, 0.1030, 0.2085], + device='cuda:1'), in_proj_covar=tensor([0.0737, 0.0666, 0.0563, 0.0657, 0.0759, 0.0620, 0.0600, 0.0614], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:00:31,744 INFO [train.py:901] (1/4) Epoch 5, batch 4050, loss[loss=0.2442, simple_loss=0.3186, pruned_loss=0.08493, over 8294.00 frames. ], tot_loss[loss=0.2903, simple_loss=0.3505, pruned_loss=0.115, over 1611393.03 frames. ], batch size: 23, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 01:00:38,671 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4592, 1.7427, 2.8636, 1.1151, 2.1349, 1.6393, 1.4016, 1.7490], + device='cuda:1'), covar=tensor([0.1438, 0.1784, 0.0556, 0.3149, 0.1230, 0.2324, 0.1430, 0.1906], + device='cuda:1'), in_proj_covar=tensor([0.0468, 0.0445, 0.0523, 0.0536, 0.0570, 0.0516, 0.0451, 0.0585], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 01:01:03,217 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 3.036e+02 3.577e+02 4.439e+02 7.437e+02, threshold=7.154e+02, percent-clipped=1.0 +2023-02-06 01:01:07,958 INFO [train.py:901] (1/4) Epoch 5, batch 4100, loss[loss=0.3363, simple_loss=0.3826, pruned_loss=0.145, over 7141.00 frames. ], tot_loss[loss=0.2907, simple_loss=0.3512, pruned_loss=0.1151, over 1612021.64 frames. ], batch size: 74, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 01:01:28,298 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 01:01:41,933 INFO [train.py:901] (1/4) Epoch 5, batch 4150, loss[loss=0.2143, simple_loss=0.2846, pruned_loss=0.07198, over 7532.00 frames. ], tot_loss[loss=0.2898, simple_loss=0.3506, pruned_loss=0.1145, over 1608997.66 frames. ], batch size: 18, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 01:01:56,690 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:13,605 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.751e+02 3.740e+02 4.679e+02 9.033e+02, threshold=7.480e+02, percent-clipped=3.0 +2023-02-06 01:02:15,177 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36529.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:17,732 INFO [train.py:901] (1/4) Epoch 5, batch 4200, loss[loss=0.3545, simple_loss=0.4041, pruned_loss=0.1525, over 8367.00 frames. ], tot_loss[loss=0.2897, simple_loss=0.3511, pruned_loss=0.1142, over 1612990.84 frames. ], batch size: 24, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:02:24,838 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:25,678 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36544.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:36,568 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3496, 1.2635, 2.3478, 1.1067, 2.0902, 2.5334, 2.4695, 2.1086], + device='cuda:1'), covar=tensor([0.0980, 0.1168, 0.0436, 0.1875, 0.0571, 0.0310, 0.0458, 0.0741], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0275, 0.0225, 0.0268, 0.0227, 0.0200, 0.0232, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 01:02:43,318 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 01:02:43,520 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36569.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:52,958 INFO [train.py:901] (1/4) Epoch 5, batch 4250, loss[loss=0.3316, simple_loss=0.3899, pruned_loss=0.1367, over 8465.00 frames. ], tot_loss[loss=0.29, simple_loss=0.3512, pruned_loss=0.1145, over 1615564.48 frames. ], batch size: 27, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:03:05,690 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 01:03:14,953 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1731, 1.1005, 3.2901, 0.9303, 2.8379, 2.7676, 2.9822, 2.9064], + device='cuda:1'), covar=tensor([0.0519, 0.3000, 0.0556, 0.2457, 0.1363, 0.0800, 0.0532, 0.0654], + device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0482, 0.0404, 0.0416, 0.0477, 0.0400, 0.0387, 0.0442], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 01:03:20,907 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:03:24,284 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36626.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:03:24,819 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 3.120e+02 3.802e+02 4.654e+02 9.583e+02, threshold=7.605e+02, percent-clipped=3.0 +2023-02-06 01:03:27,119 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:03:28,977 INFO [train.py:901] (1/4) Epoch 5, batch 4300, loss[loss=0.2757, simple_loss=0.3393, pruned_loss=0.1061, over 7964.00 frames. ], tot_loss[loss=0.2886, simple_loss=0.3498, pruned_loss=0.1137, over 1611952.20 frames. ], batch size: 21, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:03:47,173 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36658.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:04:04,991 INFO [train.py:901] (1/4) Epoch 5, batch 4350, loss[loss=0.2868, simple_loss=0.3502, pruned_loss=0.1117, over 8198.00 frames. ], tot_loss[loss=0.2881, simple_loss=0.3494, pruned_loss=0.1134, over 1615166.38 frames. ], batch size: 23, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:04:28,891 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36718.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:04:30,845 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0168, 1.6384, 4.2094, 1.7411, 3.6940, 3.6079, 3.8654, 3.7389], + device='cuda:1'), covar=tensor([0.0468, 0.3111, 0.0382, 0.2291, 0.1010, 0.0600, 0.0391, 0.0496], + device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0487, 0.0414, 0.0417, 0.0478, 0.0405, 0.0393, 0.0446], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 01:04:34,836 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 3.265e+02 4.032e+02 4.973e+02 1.053e+03, threshold=8.064e+02, percent-clipped=5.0 +2023-02-06 01:04:36,266 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 01:04:38,955 INFO [train.py:901] (1/4) Epoch 5, batch 4400, loss[loss=0.2702, simple_loss=0.3262, pruned_loss=0.1071, over 7302.00 frames. ], tot_loss[loss=0.2887, simple_loss=0.3496, pruned_loss=0.114, over 1615446.46 frames. ], batch size: 16, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:05:15,304 INFO [train.py:901] (1/4) Epoch 5, batch 4450, loss[loss=0.2707, simple_loss=0.3467, pruned_loss=0.09732, over 8191.00 frames. ], tot_loss[loss=0.2896, simple_loss=0.3497, pruned_loss=0.1147, over 1610303.17 frames. ], batch size: 23, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:05:18,076 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 01:05:43,200 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36823.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:05:45,715 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 3.029e+02 3.648e+02 4.687e+02 9.435e+02, threshold=7.296e+02, percent-clipped=4.0 +2023-02-06 01:05:49,708 INFO [train.py:901] (1/4) Epoch 5, batch 4500, loss[loss=0.2888, simple_loss=0.339, pruned_loss=0.1193, over 8037.00 frames. ], tot_loss[loss=0.2887, simple_loss=0.3493, pruned_loss=0.114, over 1609835.66 frames. ], batch size: 22, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:06:14,966 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 01:06:26,108 INFO [train.py:901] (1/4) Epoch 5, batch 4550, loss[loss=0.2796, simple_loss=0.3354, pruned_loss=0.1119, over 8082.00 frames. ], tot_loss[loss=0.2881, simple_loss=0.349, pruned_loss=0.1136, over 1610940.88 frames. ], batch size: 21, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:06:39,984 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 01:06:47,955 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36914.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:06:56,558 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 3.174e+02 3.779e+02 4.790e+02 8.988e+02, threshold=7.559e+02, percent-clipped=4.0 +2023-02-06 01:06:58,095 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36929.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:00,568 INFO [train.py:901] (1/4) Epoch 5, batch 4600, loss[loss=0.2596, simple_loss=0.3106, pruned_loss=0.1043, over 7259.00 frames. ], tot_loss[loss=0.2893, simple_loss=0.3501, pruned_loss=0.1142, over 1615269.13 frames. ], batch size: 16, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:07:04,720 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:10,000 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1476, 4.0681, 3.6813, 1.7590, 3.5990, 3.5253, 3.7582, 3.1977], + device='cuda:1'), covar=tensor([0.0940, 0.0757, 0.1153, 0.4785, 0.0899, 0.0804, 0.1624, 0.0831], + device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0292, 0.0327, 0.0410, 0.0321, 0.0273, 0.0308, 0.0253], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:07:15,528 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3023, 1.7699, 1.6947, 0.6517, 1.7304, 1.2194, 0.2389, 1.6056], + device='cuda:1'), covar=tensor([0.0169, 0.0078, 0.0066, 0.0156, 0.0094, 0.0302, 0.0267, 0.0084], + device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0225, 0.0179, 0.0264, 0.0211, 0.0358, 0.0281, 0.0257], + device='cuda:1'), out_proj_covar=tensor([1.1142e-04, 7.9431e-05, 6.1587e-05, 9.3178e-05, 7.6574e-05, 1.3819e-04, + 1.0244e-04, 9.1805e-05], device='cuda:1') +2023-02-06 01:07:23,552 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36967.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:25,631 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36970.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:29,033 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36974.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:30,328 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36976.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:35,643 INFO [train.py:901] (1/4) Epoch 5, batch 4650, loss[loss=0.2173, simple_loss=0.2804, pruned_loss=0.07711, over 7519.00 frames. ], tot_loss[loss=0.2897, simple_loss=0.3503, pruned_loss=0.1146, over 1613254.15 frames. ], batch size: 18, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:08:02,861 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:08:06,058 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 3.207e+02 3.974e+02 5.163e+02 9.904e+02, threshold=7.949e+02, percent-clipped=4.0 +2023-02-06 01:08:10,735 INFO [train.py:901] (1/4) Epoch 5, batch 4700, loss[loss=0.3055, simple_loss=0.3669, pruned_loss=0.122, over 8240.00 frames. ], tot_loss[loss=0.286, simple_loss=0.3481, pruned_loss=0.1119, over 1616198.50 frames. ], batch size: 24, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:08:16,876 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4784, 1.5603, 1.5342, 1.3703, 1.4079, 1.5176, 1.8561, 1.6514], + device='cuda:1'), covar=tensor([0.0542, 0.1210, 0.1771, 0.1449, 0.0635, 0.1457, 0.0782, 0.0575], + device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0181, 0.0222, 0.0183, 0.0132, 0.0191, 0.0147, 0.0156], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 01:08:29,891 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37062.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:08:43,608 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:08:44,127 INFO [train.py:901] (1/4) Epoch 5, batch 4750, loss[loss=0.2665, simple_loss=0.3397, pruned_loss=0.09666, over 8677.00 frames. ], tot_loss[loss=0.2874, simple_loss=0.3491, pruned_loss=0.1128, over 1616751.38 frames. ], batch size: 34, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:08:45,716 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37085.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:08:49,147 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37089.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:09:08,562 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7351, 5.8322, 5.1130, 2.0866, 5.0629, 5.4280, 5.3877, 4.8873], + device='cuda:1'), covar=tensor([0.0631, 0.0295, 0.0657, 0.4591, 0.0651, 0.0455, 0.0716, 0.0504], + device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0290, 0.0324, 0.0406, 0.0323, 0.0270, 0.0305, 0.0247], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:09:15,977 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 3.010e+02 3.846e+02 4.879e+02 1.523e+03, threshold=7.692e+02, percent-clipped=5.0 +2023-02-06 01:09:17,408 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 01:09:20,182 INFO [train.py:901] (1/4) Epoch 5, batch 4800, loss[loss=0.308, simple_loss=0.3526, pruned_loss=0.1317, over 7795.00 frames. ], tot_loss[loss=0.288, simple_loss=0.3493, pruned_loss=0.1134, over 1613702.47 frames. ], batch size: 19, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:09:20,191 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 01:09:25,059 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7394, 1.3158, 5.7948, 2.0519, 5.1550, 5.0544, 5.3918, 5.3667], + device='cuda:1'), covar=tensor([0.0403, 0.3982, 0.0227, 0.2449, 0.0847, 0.0398, 0.0307, 0.0398], + device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0486, 0.0415, 0.0413, 0.0475, 0.0399, 0.0397, 0.0443], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 01:09:40,399 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1885, 1.8400, 2.7924, 2.2987, 2.4246, 1.9168, 1.4329, 0.9261], + device='cuda:1'), covar=tensor([0.1786, 0.1855, 0.0419, 0.0905, 0.0745, 0.0933, 0.0977, 0.1802], + device='cuda:1'), in_proj_covar=tensor([0.0744, 0.0678, 0.0572, 0.0660, 0.0760, 0.0630, 0.0598, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:09:44,355 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37167.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:09:51,296 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37177.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:09:55,233 INFO [train.py:901] (1/4) Epoch 5, batch 4850, loss[loss=0.2246, simple_loss=0.2886, pruned_loss=0.0803, over 7782.00 frames. ], tot_loss[loss=0.287, simple_loss=0.348, pruned_loss=0.113, over 1612744.96 frames. ], batch size: 19, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:10:10,652 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 01:10:27,459 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.956e+02 3.581e+02 4.871e+02 1.087e+03, threshold=7.163e+02, percent-clipped=6.0 +2023-02-06 01:10:31,481 INFO [train.py:901] (1/4) Epoch 5, batch 4900, loss[loss=0.2541, simple_loss=0.3207, pruned_loss=0.09374, over 7660.00 frames. ], tot_loss[loss=0.2874, simple_loss=0.3479, pruned_loss=0.1134, over 1611239.92 frames. ], batch size: 19, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:10:59,365 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37273.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:05,611 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37282.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:11:06,080 INFO [train.py:901] (1/4) Epoch 5, batch 4950, loss[loss=0.3455, simple_loss=0.3983, pruned_loss=0.1464, over 8547.00 frames. ], tot_loss[loss=0.2879, simple_loss=0.3484, pruned_loss=0.1137, over 1613070.67 frames. ], batch size: 49, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:11:08,197 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:31,041 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37320.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:35,591 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.941e+02 3.052e+02 3.616e+02 4.696e+02 1.143e+03, threshold=7.231e+02, percent-clipped=5.0 +2023-02-06 01:11:40,307 INFO [train.py:901] (1/4) Epoch 5, batch 5000, loss[loss=0.2806, simple_loss=0.3523, pruned_loss=0.1045, over 8516.00 frames. ], tot_loss[loss=0.2894, simple_loss=0.3499, pruned_loss=0.1144, over 1615568.38 frames. ], batch size: 39, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:11:43,901 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:45,913 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37341.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:49,232 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37345.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:01,324 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37363.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:03,195 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:03,307 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:05,948 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37370.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:14,487 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37382.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:15,020 INFO [train.py:901] (1/4) Epoch 5, batch 5050, loss[loss=0.3225, simple_loss=0.3812, pruned_loss=0.1319, over 8356.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.35, pruned_loss=0.1153, over 1611753.32 frames. ], batch size: 26, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:12:18,411 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:22,598 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 01:12:24,797 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-02-06 01:12:44,491 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.841e+02 3.496e+02 4.122e+02 5.072e+02 9.522e+02, threshold=8.245e+02, percent-clipped=6.0 +2023-02-06 01:12:48,504 INFO [train.py:901] (1/4) Epoch 5, batch 5100, loss[loss=0.2689, simple_loss=0.3496, pruned_loss=0.09412, over 8512.00 frames. ], tot_loss[loss=0.2894, simple_loss=0.3492, pruned_loss=0.1148, over 1608395.07 frames. ], batch size: 26, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:12:48,513 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 01:12:48,708 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37433.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:12:49,923 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:13:01,262 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4765, 2.0818, 3.6320, 1.0504, 2.3630, 1.7489, 1.5374, 2.0909], + device='cuda:1'), covar=tensor([0.1454, 0.1529, 0.0623, 0.3093, 0.1351, 0.2444, 0.1411, 0.2101], + device='cuda:1'), in_proj_covar=tensor([0.0464, 0.0438, 0.0514, 0.0532, 0.0570, 0.0511, 0.0442, 0.0580], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 01:13:06,543 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37458.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:13:07,425 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 01:13:22,241 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37481.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:13:23,387 INFO [train.py:901] (1/4) Epoch 5, batch 5150, loss[loss=0.3371, simple_loss=0.3804, pruned_loss=0.1469, over 6824.00 frames. ], tot_loss[loss=0.2884, simple_loss=0.3483, pruned_loss=0.1142, over 1608405.38 frames. ], batch size: 71, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:13:53,503 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 3.138e+02 3.879e+02 5.454e+02 1.167e+03, threshold=7.757e+02, percent-clipped=4.0 +2023-02-06 01:13:57,569 INFO [train.py:901] (1/4) Epoch 5, batch 5200, loss[loss=0.2721, simple_loss=0.3432, pruned_loss=0.1006, over 8585.00 frames. ], tot_loss[loss=0.2882, simple_loss=0.3486, pruned_loss=0.1139, over 1609859.10 frames. ], batch size: 34, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:14:01,118 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37538.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:14:19,383 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37563.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:14:27,418 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.15 vs. limit=5.0 +2023-02-06 01:14:33,081 INFO [train.py:901] (1/4) Epoch 5, batch 5250, loss[loss=0.2951, simple_loss=0.3495, pruned_loss=0.1204, over 7792.00 frames. ], tot_loss[loss=0.29, simple_loss=0.3502, pruned_loss=0.1149, over 1612732.15 frames. ], batch size: 19, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:14:45,236 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 01:15:03,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 3.205e+02 3.781e+02 5.298e+02 9.083e+02, threshold=7.562e+02, percent-clipped=4.0 +2023-02-06 01:15:05,549 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:15:07,537 INFO [train.py:901] (1/4) Epoch 5, batch 5300, loss[loss=0.2868, simple_loss=0.3464, pruned_loss=0.1136, over 8099.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3505, pruned_loss=0.1149, over 1617389.89 frames. ], batch size: 23, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:15:15,192 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:15:28,082 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3635, 1.6865, 2.8652, 1.0883, 2.0858, 1.6126, 1.4619, 1.6505], + device='cuda:1'), covar=tensor([0.1408, 0.1633, 0.0558, 0.2920, 0.1171, 0.2162, 0.1360, 0.1798], + device='cuda:1'), in_proj_covar=tensor([0.0459, 0.0438, 0.0506, 0.0530, 0.0561, 0.0504, 0.0440, 0.0576], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 01:15:32,586 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37669.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:15:42,454 INFO [train.py:901] (1/4) Epoch 5, batch 5350, loss[loss=0.3417, simple_loss=0.3874, pruned_loss=0.148, over 6805.00 frames. ], tot_loss[loss=0.2927, simple_loss=0.353, pruned_loss=0.1162, over 1620855.40 frames. ], batch size: 71, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:15:47,813 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:05,041 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37716.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:11,801 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37726.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:12,347 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.725e+02 3.472e+02 4.206e+02 5.536e+02 1.524e+03, threshold=8.412e+02, percent-clipped=7.0 +2023-02-06 01:16:17,026 INFO [train.py:901] (1/4) Epoch 5, batch 5400, loss[loss=0.2352, simple_loss=0.2966, pruned_loss=0.08686, over 7804.00 frames. ], tot_loss[loss=0.2932, simple_loss=0.3534, pruned_loss=0.1166, over 1620048.64 frames. ], batch size: 19, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:16:17,260 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8955, 2.1367, 4.1062, 1.2619, 2.9111, 2.2861, 1.7336, 2.4980], + device='cuda:1'), covar=tensor([0.1275, 0.1772, 0.0645, 0.3197, 0.1221, 0.2109, 0.1417, 0.2144], + device='cuda:1'), in_proj_covar=tensor([0.0461, 0.0440, 0.0515, 0.0535, 0.0570, 0.0509, 0.0442, 0.0580], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 01:16:19,896 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:25,204 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:36,760 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37762.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:40,183 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1454, 1.6327, 2.7435, 2.1482, 2.4107, 1.7783, 1.3751, 0.9416], + device='cuda:1'), covar=tensor([0.1769, 0.2113, 0.0458, 0.1066, 0.0844, 0.1016, 0.1082, 0.1990], + device='cuda:1'), in_proj_covar=tensor([0.0754, 0.0680, 0.0573, 0.0663, 0.0777, 0.0631, 0.0609, 0.0629], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:16:50,828 INFO [train.py:901] (1/4) Epoch 5, batch 5450, loss[loss=0.2967, simple_loss=0.3482, pruned_loss=0.1226, over 8087.00 frames. ], tot_loss[loss=0.2921, simple_loss=0.3521, pruned_loss=0.116, over 1615584.06 frames. ], batch size: 21, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:16:54,347 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9989, 1.3539, 3.3091, 1.1476, 2.1660, 3.6475, 3.6142, 3.0802], + device='cuda:1'), covar=tensor([0.1060, 0.1643, 0.0431, 0.2199, 0.0960, 0.0264, 0.0340, 0.0709], + device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0273, 0.0225, 0.0262, 0.0224, 0.0200, 0.0230, 0.0279], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 01:17:22,462 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.936e+02 3.882e+02 5.021e+02 1.156e+03, threshold=7.764e+02, percent-clipped=3.0 +2023-02-06 01:17:26,495 INFO [train.py:901] (1/4) Epoch 5, batch 5500, loss[loss=0.3106, simple_loss=0.3628, pruned_loss=0.1291, over 8327.00 frames. ], tot_loss[loss=0.2915, simple_loss=0.3514, pruned_loss=0.1158, over 1617304.87 frames. ], batch size: 26, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:17:30,475 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 01:17:32,001 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:18:00,740 INFO [train.py:901] (1/4) Epoch 5, batch 5550, loss[loss=0.3148, simple_loss=0.3849, pruned_loss=0.1223, over 8198.00 frames. ], tot_loss[loss=0.2921, simple_loss=0.3526, pruned_loss=0.1158, over 1616894.89 frames. ], batch size: 23, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:18:32,130 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 3.124e+02 3.941e+02 5.093e+02 9.977e+02, threshold=7.882e+02, percent-clipped=4.0 +2023-02-06 01:18:36,200 INFO [train.py:901] (1/4) Epoch 5, batch 5600, loss[loss=0.2761, simple_loss=0.345, pruned_loss=0.1036, over 8187.00 frames. ], tot_loss[loss=0.29, simple_loss=0.3503, pruned_loss=0.1148, over 1613186.20 frames. ], batch size: 23, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:19:01,903 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5297, 4.3970, 3.9953, 2.3803, 3.9403, 4.0637, 4.1333, 3.6439], + device='cuda:1'), covar=tensor([0.0895, 0.0605, 0.0858, 0.3801, 0.0754, 0.0763, 0.1124, 0.0825], + device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0290, 0.0316, 0.0405, 0.0323, 0.0270, 0.0299, 0.0251], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:19:04,040 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5839, 3.0500, 3.0195, 2.0772, 1.6112, 3.1086, 0.5303, 1.8793], + device='cuda:1'), covar=tensor([0.4361, 0.1831, 0.0982, 0.4012, 0.6655, 0.0671, 0.6758, 0.2401], + device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0121, 0.0079, 0.0172, 0.0208, 0.0079, 0.0146, 0.0124], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:19:10,753 INFO [train.py:901] (1/4) Epoch 5, batch 5650, loss[loss=0.2689, simple_loss=0.3345, pruned_loss=0.1017, over 7963.00 frames. ], tot_loss[loss=0.2899, simple_loss=0.3503, pruned_loss=0.1147, over 1615469.98 frames. ], batch size: 21, lr: 1.47e-02, grad_scale: 4.0 +2023-02-06 01:19:20,216 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37997.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:19:23,907 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38001.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:19:34,396 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 01:19:41,409 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38026.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:19:42,540 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 3.005e+02 3.717e+02 4.758e+02 1.120e+03, threshold=7.434e+02, percent-clipped=3.0 +2023-02-06 01:19:45,907 INFO [train.py:901] (1/4) Epoch 5, batch 5700, loss[loss=0.3328, simple_loss=0.3852, pruned_loss=0.1402, over 8685.00 frames. ], tot_loss[loss=0.291, simple_loss=0.3511, pruned_loss=0.1155, over 1614634.39 frames. ], batch size: 34, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:19:55,551 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4025, 1.7045, 2.8403, 1.1119, 2.0626, 1.7906, 1.3327, 1.7540], + device='cuda:1'), covar=tensor([0.1521, 0.1764, 0.0672, 0.3194, 0.1291, 0.2279, 0.1664, 0.1735], + device='cuda:1'), in_proj_covar=tensor([0.0466, 0.0448, 0.0530, 0.0537, 0.0577, 0.0518, 0.0448, 0.0586], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 01:20:07,611 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0066, 2.2799, 2.8899, 0.9299, 2.9884, 1.9318, 1.3132, 1.5912], + device='cuda:1'), covar=tensor([0.0286, 0.0136, 0.0116, 0.0276, 0.0178, 0.0306, 0.0369, 0.0193], + device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0224, 0.0183, 0.0268, 0.0214, 0.0360, 0.0281, 0.0262], + device='cuda:1'), out_proj_covar=tensor([1.1335e-04, 7.7842e-05, 6.2803e-05, 9.3467e-05, 7.6265e-05, 1.3717e-04, + 1.0081e-04, 9.1940e-05], device='cuda:1') +2023-02-06 01:20:20,728 INFO [train.py:901] (1/4) Epoch 5, batch 5750, loss[loss=0.2211, simple_loss=0.2983, pruned_loss=0.07193, over 7935.00 frames. ], tot_loss[loss=0.2914, simple_loss=0.3515, pruned_loss=0.1156, over 1612192.81 frames. ], batch size: 20, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:20:30,232 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38097.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:20:37,218 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 01:20:46,793 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38122.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:20:50,583 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 3.053e+02 3.876e+02 4.925e+02 1.023e+03, threshold=7.752e+02, percent-clipped=4.0 +2023-02-06 01:20:54,530 INFO [train.py:901] (1/4) Epoch 5, batch 5800, loss[loss=0.3103, simple_loss=0.3732, pruned_loss=0.1237, over 8200.00 frames. ], tot_loss[loss=0.2918, simple_loss=0.3524, pruned_loss=0.1156, over 1614187.60 frames. ], batch size: 23, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:20:59,287 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2123, 1.4540, 4.0611, 1.8231, 2.4428, 4.6068, 4.5838, 4.0864], + device='cuda:1'), covar=tensor([0.1057, 0.1545, 0.0292, 0.1808, 0.0819, 0.0199, 0.0261, 0.0501], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0269, 0.0225, 0.0263, 0.0224, 0.0200, 0.0232, 0.0277], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 01:21:25,613 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3276, 1.6320, 1.6654, 1.2869, 0.8800, 1.6708, 0.1102, 1.1107], + device='cuda:1'), covar=tensor([0.3682, 0.1970, 0.1066, 0.2061, 0.5300, 0.1112, 0.5127, 0.2090], + device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0121, 0.0081, 0.0169, 0.0210, 0.0081, 0.0145, 0.0125], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:21:26,249 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8828, 1.2000, 1.4163, 1.0654, 1.1072, 1.3216, 1.5029, 1.5003], + device='cuda:1'), covar=tensor([0.0685, 0.1459, 0.2130, 0.1696, 0.0673, 0.1836, 0.0848, 0.0629], + device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0178, 0.0218, 0.0182, 0.0130, 0.0189, 0.0144, 0.0153], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 01:21:29,423 INFO [train.py:901] (1/4) Epoch 5, batch 5850, loss[loss=0.2592, simple_loss=0.3316, pruned_loss=0.09338, over 8257.00 frames. ], tot_loss[loss=0.2906, simple_loss=0.3517, pruned_loss=0.1148, over 1612223.37 frames. ], batch size: 24, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:22:01,295 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 3.016e+02 3.759e+02 4.889e+02 1.185e+03, threshold=7.518e+02, percent-clipped=2.0 +2023-02-06 01:22:04,825 INFO [train.py:901] (1/4) Epoch 5, batch 5900, loss[loss=0.2834, simple_loss=0.3591, pruned_loss=0.1039, over 8247.00 frames. ], tot_loss[loss=0.2893, simple_loss=0.3508, pruned_loss=0.1139, over 1612957.78 frames. ], batch size: 24, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:22:25,289 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5979, 4.6196, 3.9406, 1.9169, 4.0126, 3.9781, 4.2545, 3.6383], + device='cuda:1'), covar=tensor([0.0772, 0.0530, 0.1031, 0.4718, 0.0870, 0.0924, 0.1090, 0.0772], + device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0285, 0.0319, 0.0404, 0.0314, 0.0268, 0.0298, 0.0251], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:22:40,888 INFO [train.py:901] (1/4) Epoch 5, batch 5950, loss[loss=0.3138, simple_loss=0.3736, pruned_loss=0.127, over 8547.00 frames. ], tot_loss[loss=0.2904, simple_loss=0.3518, pruned_loss=0.1145, over 1613653.82 frames. ], batch size: 49, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:23:12,028 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 3.191e+02 3.790e+02 5.332e+02 1.075e+03, threshold=7.580e+02, percent-clipped=7.0 +2023-02-06 01:23:15,487 INFO [train.py:901] (1/4) Epoch 5, batch 6000, loss[loss=0.2611, simple_loss=0.3224, pruned_loss=0.09988, over 8239.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3514, pruned_loss=0.1145, over 1613781.17 frames. ], batch size: 22, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:23:15,487 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 01:23:28,278 INFO [train.py:935] (1/4) Epoch 5, validation: loss=0.2196, simple_loss=0.3162, pruned_loss=0.06146, over 944034.00 frames. +2023-02-06 01:23:28,279 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-06 01:23:33,799 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=38341.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:23:58,476 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38378.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:23:58,573 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4717, 1.7854, 3.0798, 1.0140, 2.3020, 1.7680, 1.5088, 1.8062], + device='cuda:1'), covar=tensor([0.1418, 0.1665, 0.0628, 0.3140, 0.1291, 0.2264, 0.1378, 0.2140], + device='cuda:1'), in_proj_covar=tensor([0.0465, 0.0445, 0.0523, 0.0536, 0.0578, 0.0519, 0.0443, 0.0594], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 01:24:01,364 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 01:24:01,709 INFO [train.py:901] (1/4) Epoch 5, batch 6050, loss[loss=0.2522, simple_loss=0.3353, pruned_loss=0.08457, over 8481.00 frames. ], tot_loss[loss=0.2898, simple_loss=0.3514, pruned_loss=0.1141, over 1615432.89 frames. ], batch size: 27, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:24:33,765 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 3.108e+02 3.868e+02 4.827e+02 8.119e+02, threshold=7.737e+02, percent-clipped=1.0 +2023-02-06 01:24:37,063 INFO [train.py:901] (1/4) Epoch 5, batch 6100, loss[loss=0.2681, simple_loss=0.3354, pruned_loss=0.1004, over 8030.00 frames. ], tot_loss[loss=0.2897, simple_loss=0.3512, pruned_loss=0.1141, over 1615452.54 frames. ], batch size: 22, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:24:53,394 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38456.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:25:09,663 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 01:25:10,932 INFO [train.py:901] (1/4) Epoch 5, batch 6150, loss[loss=0.2583, simple_loss=0.3188, pruned_loss=0.09888, over 7929.00 frames. ], tot_loss[loss=0.2887, simple_loss=0.3502, pruned_loss=0.1136, over 1611638.73 frames. ], batch size: 20, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:25:42,288 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.784e+02 3.089e+02 4.008e+02 5.119e+02 1.011e+03, threshold=8.016e+02, percent-clipped=7.0 +2023-02-06 01:25:45,625 INFO [train.py:901] (1/4) Epoch 5, batch 6200, loss[loss=0.3564, simple_loss=0.3947, pruned_loss=0.1591, over 8520.00 frames. ], tot_loss[loss=0.2878, simple_loss=0.3486, pruned_loss=0.1135, over 1607090.10 frames. ], batch size: 26, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:25:45,970 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.80 vs. limit=5.0 +2023-02-06 01:26:20,120 INFO [train.py:901] (1/4) Epoch 5, batch 6250, loss[loss=0.3296, simple_loss=0.3911, pruned_loss=0.1341, over 8507.00 frames. ], tot_loss[loss=0.287, simple_loss=0.348, pruned_loss=0.113, over 1607273.19 frames. ], batch size: 26, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:26:51,177 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.140e+02 3.239e+02 3.994e+02 4.997e+02 1.061e+03, threshold=7.988e+02, percent-clipped=3.0 +2023-02-06 01:26:54,599 INFO [train.py:901] (1/4) Epoch 5, batch 6300, loss[loss=0.335, simple_loss=0.3889, pruned_loss=0.1406, over 8255.00 frames. ], tot_loss[loss=0.2874, simple_loss=0.3482, pruned_loss=0.1133, over 1606443.60 frames. ], batch size: 24, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:27:29,613 INFO [train.py:901] (1/4) Epoch 5, batch 6350, loss[loss=0.2823, simple_loss=0.3389, pruned_loss=0.1129, over 7533.00 frames. ], tot_loss[loss=0.2853, simple_loss=0.3462, pruned_loss=0.1122, over 1600498.30 frames. ], batch size: 18, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:27:49,913 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38712.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:27:56,421 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=38722.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:28:00,318 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.862e+02 3.826e+02 4.732e+02 1.596e+03, threshold=7.652e+02, percent-clipped=5.0 +2023-02-06 01:28:03,602 INFO [train.py:901] (1/4) Epoch 5, batch 6400, loss[loss=0.2555, simple_loss=0.3169, pruned_loss=0.09708, over 7659.00 frames. ], tot_loss[loss=0.285, simple_loss=0.3467, pruned_loss=0.1116, over 1606835.37 frames. ], batch size: 19, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:28:06,533 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:28:12,650 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 01:28:38,882 INFO [train.py:901] (1/4) Epoch 5, batch 6450, loss[loss=0.2853, simple_loss=0.3485, pruned_loss=0.111, over 8453.00 frames. ], tot_loss[loss=0.2843, simple_loss=0.3462, pruned_loss=0.1112, over 1606896.47 frames. ], batch size: 27, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:28:54,128 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5489, 1.7374, 2.7666, 1.2883, 2.1667, 1.7783, 1.6185, 1.9775], + device='cuda:1'), covar=tensor([0.1086, 0.1471, 0.0499, 0.2378, 0.1034, 0.1704, 0.1142, 0.1476], + device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0454, 0.0524, 0.0541, 0.0578, 0.0529, 0.0450, 0.0595], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 01:29:09,808 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.470e+02 3.536e+02 4.141e+02 5.010e+02 9.096e+02, threshold=8.281e+02, percent-clipped=4.0 +2023-02-06 01:29:13,125 INFO [train.py:901] (1/4) Epoch 5, batch 6500, loss[loss=0.2782, simple_loss=0.3459, pruned_loss=0.1053, over 8135.00 frames. ], tot_loss[loss=0.2856, simple_loss=0.3472, pruned_loss=0.112, over 1605443.04 frames. ], batch size: 22, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:29:16,032 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38837.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:29:18,637 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:29:48,026 INFO [train.py:901] (1/4) Epoch 5, batch 6550, loss[loss=0.3243, simple_loss=0.3781, pruned_loss=0.1352, over 8566.00 frames. ], tot_loss[loss=0.2853, simple_loss=0.3469, pruned_loss=0.1118, over 1602904.06 frames. ], batch size: 49, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:30:19,275 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.070e+02 3.239e+02 3.737e+02 4.952e+02 1.438e+03, threshold=7.474e+02, percent-clipped=4.0 +2023-02-06 01:30:22,018 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 01:30:22,677 INFO [train.py:901] (1/4) Epoch 5, batch 6600, loss[loss=0.248, simple_loss=0.3119, pruned_loss=0.09205, over 7799.00 frames. ], tot_loss[loss=0.2865, simple_loss=0.3477, pruned_loss=0.1126, over 1606366.61 frames. ], batch size: 20, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:30:39,812 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 01:30:41,192 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6674, 4.6633, 4.1408, 1.8361, 4.0879, 4.1757, 4.2840, 3.8642], + device='cuda:1'), covar=tensor([0.0766, 0.0553, 0.1138, 0.5246, 0.0789, 0.0601, 0.1211, 0.0735], + device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0300, 0.0327, 0.0414, 0.0321, 0.0280, 0.0305, 0.0256], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:30:58,095 INFO [train.py:901] (1/4) Epoch 5, batch 6650, loss[loss=0.2766, simple_loss=0.3476, pruned_loss=0.1028, over 8502.00 frames. ], tot_loss[loss=0.2867, simple_loss=0.348, pruned_loss=0.1128, over 1608337.82 frames. ], batch size: 26, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:31:08,532 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-02-06 01:31:10,555 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 01:31:29,862 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 3.068e+02 3.660e+02 4.252e+02 1.265e+03, threshold=7.321e+02, percent-clipped=3.0 +2023-02-06 01:31:33,329 INFO [train.py:901] (1/4) Epoch 5, batch 6700, loss[loss=0.282, simple_loss=0.3559, pruned_loss=0.1041, over 8480.00 frames. ], tot_loss[loss=0.2867, simple_loss=0.3482, pruned_loss=0.1126, over 1610436.77 frames. ], batch size: 25, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:31:56,923 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-02-06 01:32:08,098 INFO [train.py:901] (1/4) Epoch 5, batch 6750, loss[loss=0.2627, simple_loss=0.3333, pruned_loss=0.09601, over 8105.00 frames. ], tot_loss[loss=0.2854, simple_loss=0.3477, pruned_loss=0.1115, over 1613596.92 frames. ], batch size: 23, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:32:15,834 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39093.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:32:33,090 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39118.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:32:40,327 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 2.802e+02 3.437e+02 4.300e+02 8.945e+02, threshold=6.874e+02, percent-clipped=2.0 +2023-02-06 01:32:43,709 INFO [train.py:901] (1/4) Epoch 5, batch 6800, loss[loss=0.2272, simple_loss=0.2935, pruned_loss=0.08045, over 7698.00 frames. ], tot_loss[loss=0.2841, simple_loss=0.3462, pruned_loss=0.111, over 1605710.65 frames. ], batch size: 18, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:32:54,744 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 01:33:04,419 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3909, 2.8995, 2.0380, 2.2742, 2.3359, 1.8264, 2.3023, 2.2878], + device='cuda:1'), covar=tensor([0.1350, 0.0219, 0.0835, 0.0640, 0.0612, 0.1074, 0.0733, 0.0965], + device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0242, 0.0303, 0.0300, 0.0315, 0.0312, 0.0334, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 01:33:10,331 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8484, 5.8448, 5.1795, 2.0401, 5.2037, 5.4688, 5.4850, 5.1800], + device='cuda:1'), covar=tensor([0.0496, 0.0400, 0.0734, 0.4494, 0.0538, 0.0594, 0.0845, 0.0537], + device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0292, 0.0320, 0.0404, 0.0311, 0.0280, 0.0304, 0.0252], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:33:17,612 INFO [train.py:901] (1/4) Epoch 5, batch 6850, loss[loss=0.2569, simple_loss=0.3105, pruned_loss=0.1017, over 7535.00 frames. ], tot_loss[loss=0.2848, simple_loss=0.3466, pruned_loss=0.1115, over 1608417.86 frames. ], batch size: 18, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:33:19,093 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:33:31,228 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39203.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:33:43,502 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 01:33:48,721 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 3.284e+02 3.960e+02 5.468e+02 1.321e+03, threshold=7.919e+02, percent-clipped=11.0 +2023-02-06 01:33:52,161 INFO [train.py:901] (1/4) Epoch 5, batch 6900, loss[loss=0.3038, simple_loss=0.3638, pruned_loss=0.1219, over 8073.00 frames. ], tot_loss[loss=0.2858, simple_loss=0.3477, pruned_loss=0.1119, over 1610867.45 frames. ], batch size: 21, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:34:26,649 INFO [train.py:901] (1/4) Epoch 5, batch 6950, loss[loss=0.2547, simple_loss=0.3356, pruned_loss=0.08694, over 8461.00 frames. ], tot_loss[loss=0.2864, simple_loss=0.3482, pruned_loss=0.1123, over 1611790.13 frames. ], batch size: 25, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:34:38,150 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39300.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:34:39,401 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39302.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:34:49,528 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 01:34:58,303 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 3.231e+02 3.801e+02 5.196e+02 1.038e+03, threshold=7.603e+02, percent-clipped=4.0 +2023-02-06 01:35:01,665 INFO [train.py:901] (1/4) Epoch 5, batch 7000, loss[loss=0.2455, simple_loss=0.3148, pruned_loss=0.08808, over 8091.00 frames. ], tot_loss[loss=0.285, simple_loss=0.3474, pruned_loss=0.1113, over 1614099.48 frames. ], batch size: 21, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:35:06,877 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 01:35:34,736 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5074, 2.9094, 3.0922, 1.7544, 1.5423, 2.7769, 0.5486, 1.8616], + device='cuda:1'), covar=tensor([0.4420, 0.1674, 0.0718, 0.3500, 0.5483, 0.0843, 0.5902, 0.2435], + device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0128, 0.0082, 0.0175, 0.0217, 0.0082, 0.0143, 0.0130], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:35:35,823 INFO [train.py:901] (1/4) Epoch 5, batch 7050, loss[loss=0.2413, simple_loss=0.3049, pruned_loss=0.08887, over 7247.00 frames. ], tot_loss[loss=0.284, simple_loss=0.3466, pruned_loss=0.1107, over 1614672.79 frames. ], batch size: 16, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:35:40,290 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 01:36:04,433 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4085, 1.8578, 3.1558, 1.1104, 2.2372, 1.8449, 1.4896, 1.8704], + device='cuda:1'), covar=tensor([0.1533, 0.1788, 0.0543, 0.3172, 0.1214, 0.2257, 0.1517, 0.2161], + device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0449, 0.0514, 0.0522, 0.0564, 0.0505, 0.0440, 0.0582], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 01:36:06,897 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 2.867e+02 3.538e+02 4.706e+02 1.662e+03, threshold=7.075e+02, percent-clipped=2.0 +2023-02-06 01:36:10,287 INFO [train.py:901] (1/4) Epoch 5, batch 7100, loss[loss=0.2581, simple_loss=0.3349, pruned_loss=0.09063, over 8145.00 frames. ], tot_loss[loss=0.284, simple_loss=0.3469, pruned_loss=0.1106, over 1617120.76 frames. ], batch size: 22, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:36:10,520 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0520, 2.5508, 3.2013, 1.1135, 2.9130, 1.9039, 1.3748, 1.7928], + device='cuda:1'), covar=tensor([0.0306, 0.0133, 0.0097, 0.0273, 0.0240, 0.0363, 0.0377, 0.0191], + device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0229, 0.0195, 0.0277, 0.0224, 0.0374, 0.0291, 0.0270], + device='cuda:1'), out_proj_covar=tensor([1.1320e-04, 7.8686e-05, 6.6479e-05, 9.5059e-05, 7.8630e-05, 1.4069e-04, + 1.0271e-04, 9.3335e-05], device='cuda:1') +2023-02-06 01:36:44,804 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39481.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:36:46,001 INFO [train.py:901] (1/4) Epoch 5, batch 7150, loss[loss=0.307, simple_loss=0.3689, pruned_loss=0.1225, over 8726.00 frames. ], tot_loss[loss=0.2842, simple_loss=0.3473, pruned_loss=0.1106, over 1617978.95 frames. ], batch size: 34, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:36:50,219 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0332, 1.3013, 4.1777, 1.5731, 3.5797, 3.4149, 3.7319, 3.6637], + device='cuda:1'), covar=tensor([0.0441, 0.3799, 0.0479, 0.2637, 0.1200, 0.0693, 0.0509, 0.0606], + device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0500, 0.0439, 0.0433, 0.0502, 0.0419, 0.0413, 0.0466], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 01:37:17,188 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 2.912e+02 3.907e+02 4.774e+02 1.202e+03, threshold=7.813e+02, percent-clipped=7.0 +2023-02-06 01:37:18,078 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6495, 1.6222, 4.7946, 1.7027, 4.1797, 3.9700, 4.3308, 4.1751], + device='cuda:1'), covar=tensor([0.0332, 0.3125, 0.0296, 0.2350, 0.0975, 0.0474, 0.0364, 0.0460], + device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0492, 0.0435, 0.0431, 0.0497, 0.0414, 0.0408, 0.0462], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 01:37:20,770 INFO [train.py:901] (1/4) Epoch 5, batch 7200, loss[loss=0.2479, simple_loss=0.3228, pruned_loss=0.08647, over 8130.00 frames. ], tot_loss[loss=0.2861, simple_loss=0.3491, pruned_loss=0.1115, over 1622025.62 frames. ], batch size: 22, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:37:21,582 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5540, 4.5235, 4.0784, 1.8527, 3.9734, 4.0861, 4.2149, 3.7080], + device='cuda:1'), covar=tensor([0.0721, 0.0566, 0.0881, 0.4503, 0.0630, 0.0610, 0.1170, 0.0713], + device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0288, 0.0317, 0.0403, 0.0306, 0.0275, 0.0301, 0.0249], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:37:21,686 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4818, 1.7722, 2.1636, 0.9972, 2.1224, 1.4129, 0.6507, 1.6596], + device='cuda:1'), covar=tensor([0.0281, 0.0138, 0.0104, 0.0229, 0.0143, 0.0361, 0.0325, 0.0129], + device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0226, 0.0192, 0.0274, 0.0222, 0.0370, 0.0289, 0.0267], + device='cuda:1'), out_proj_covar=tensor([1.1218e-04, 7.7630e-05, 6.5339e-05, 9.3752e-05, 7.7743e-05, 1.3891e-04, + 1.0154e-04, 9.2352e-05], device='cuda:1') +2023-02-06 01:37:30,602 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39547.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:37:37,515 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39556.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:37:55,140 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39581.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:37:56,298 INFO [train.py:901] (1/4) Epoch 5, batch 7250, loss[loss=0.2585, simple_loss=0.3223, pruned_loss=0.09733, over 7644.00 frames. ], tot_loss[loss=0.2849, simple_loss=0.3479, pruned_loss=0.1109, over 1621789.10 frames. ], batch size: 19, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:38:27,208 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 2.862e+02 3.679e+02 5.056e+02 1.142e+03, threshold=7.358e+02, percent-clipped=8.0 +2023-02-06 01:38:30,500 INFO [train.py:901] (1/4) Epoch 5, batch 7300, loss[loss=0.2913, simple_loss=0.3438, pruned_loss=0.1194, over 7439.00 frames. ], tot_loss[loss=0.2862, simple_loss=0.3486, pruned_loss=0.1119, over 1620717.40 frames. ], batch size: 17, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:38:39,559 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39646.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:38:47,105 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39657.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:38:50,512 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39662.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:39:05,556 INFO [train.py:901] (1/4) Epoch 5, batch 7350, loss[loss=0.2685, simple_loss=0.3324, pruned_loss=0.1024, over 8357.00 frames. ], tot_loss[loss=0.2866, simple_loss=0.3484, pruned_loss=0.1125, over 1615177.25 frames. ], batch size: 24, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:39:33,447 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 01:39:36,159 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.150e+02 2.846e+02 3.982e+02 4.999e+02 1.878e+03, threshold=7.964e+02, percent-clipped=11.0 +2023-02-06 01:39:39,633 INFO [train.py:901] (1/4) Epoch 5, batch 7400, loss[loss=0.3019, simple_loss=0.3648, pruned_loss=0.1194, over 8605.00 frames. ], tot_loss[loss=0.2886, simple_loss=0.3501, pruned_loss=0.1135, over 1619484.09 frames. ], batch size: 49, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:39:52,995 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 01:39:59,153 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:40:01,203 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1329, 2.4334, 3.0940, 1.2397, 3.2125, 2.1396, 1.5200, 1.7748], + device='cuda:1'), covar=tensor([0.0262, 0.0115, 0.0111, 0.0260, 0.0116, 0.0308, 0.0365, 0.0185], + device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0223, 0.0192, 0.0276, 0.0221, 0.0370, 0.0289, 0.0266], + device='cuda:1'), out_proj_covar=tensor([1.1039e-04, 7.6072e-05, 6.5133e-05, 9.4686e-05, 7.7486e-05, 1.3836e-04, + 1.0171e-04, 9.1758e-05], device='cuda:1') +2023-02-06 01:40:05,265 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1139, 1.6622, 4.1853, 1.9392, 2.3620, 4.9006, 4.7569, 4.3198], + device='cuda:1'), covar=tensor([0.1078, 0.1462, 0.0308, 0.1913, 0.0784, 0.0189, 0.0299, 0.0475], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0263, 0.0222, 0.0259, 0.0220, 0.0198, 0.0230, 0.0272], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 01:40:13,746 INFO [train.py:901] (1/4) Epoch 5, batch 7450, loss[loss=0.2412, simple_loss=0.2963, pruned_loss=0.0931, over 7244.00 frames. ], tot_loss[loss=0.2879, simple_loss=0.35, pruned_loss=0.1129, over 1620937.37 frames. ], batch size: 16, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:40:15,955 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:40:32,187 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 01:40:43,637 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39825.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:40:43,747 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1809, 1.6282, 1.6281, 1.5621, 1.6982, 1.6154, 2.5570, 2.2648], + device='cuda:1'), covar=tensor([0.0484, 0.1257, 0.1654, 0.1313, 0.0539, 0.1517, 0.0603, 0.0490], + device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0177, 0.0217, 0.0181, 0.0128, 0.0187, 0.0142, 0.0150], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 01:40:45,625 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.116e+02 3.779e+02 4.440e+02 1.107e+03, threshold=7.558e+02, percent-clipped=3.0 +2023-02-06 01:40:49,055 INFO [train.py:901] (1/4) Epoch 5, batch 7500, loss[loss=0.3468, simple_loss=0.3904, pruned_loss=0.1516, over 7017.00 frames. ], tot_loss[loss=0.2883, simple_loss=0.3499, pruned_loss=0.1133, over 1619170.92 frames. ], batch size: 71, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:40:49,187 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:41:21,798 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7710, 1.5537, 3.2468, 1.3164, 2.0802, 3.7596, 3.6634, 3.1478], + device='cuda:1'), covar=tensor([0.1085, 0.1400, 0.0355, 0.2055, 0.0837, 0.0216, 0.0344, 0.0575], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0265, 0.0224, 0.0263, 0.0224, 0.0202, 0.0234, 0.0278], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 01:41:22,994 INFO [train.py:901] (1/4) Epoch 5, batch 7550, loss[loss=0.2732, simple_loss=0.3456, pruned_loss=0.1004, over 8110.00 frames. ], tot_loss[loss=0.2875, simple_loss=0.3495, pruned_loss=0.1128, over 1615553.61 frames. ], batch size: 23, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:41:48,014 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39918.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:41:54,548 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.961e+02 3.144e+02 3.978e+02 5.379e+02 1.554e+03, threshold=7.957e+02, percent-clipped=6.0 +2023-02-06 01:41:57,850 INFO [train.py:901] (1/4) Epoch 5, batch 7600, loss[loss=0.3406, simple_loss=0.3826, pruned_loss=0.1493, over 7333.00 frames. ], tot_loss[loss=0.2868, simple_loss=0.3491, pruned_loss=0.1123, over 1616095.80 frames. ], batch size: 71, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:42:02,463 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:42:04,536 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39943.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:42:19,502 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4354, 2.1581, 1.5726, 1.8930, 1.8258, 1.4447, 1.7772, 1.8643], + device='cuda:1'), covar=tensor([0.0728, 0.0251, 0.0636, 0.0341, 0.0396, 0.0765, 0.0523, 0.0517], + device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0241, 0.0316, 0.0301, 0.0314, 0.0313, 0.0334, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 01:42:32,756 INFO [train.py:901] (1/4) Epoch 5, batch 7650, loss[loss=0.2915, simple_loss=0.3542, pruned_loss=0.1144, over 8369.00 frames. ], tot_loss[loss=0.2877, simple_loss=0.3495, pruned_loss=0.113, over 1617435.92 frames. ], batch size: 49, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:42:43,218 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.69 vs. limit=5.0 +2023-02-06 01:42:45,885 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40001.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:42:57,000 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40017.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:43:05,594 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.281e+02 3.028e+02 3.689e+02 4.703e+02 1.290e+03, threshold=7.379e+02, percent-clipped=1.0 +2023-02-06 01:43:08,877 INFO [train.py:901] (1/4) Epoch 5, batch 7700, loss[loss=0.3168, simple_loss=0.365, pruned_loss=0.1343, over 8339.00 frames. ], tot_loss[loss=0.2882, simple_loss=0.3496, pruned_loss=0.1134, over 1617231.71 frames. ], batch size: 48, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:43:15,432 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:43:44,266 INFO [train.py:901] (1/4) Epoch 5, batch 7750, loss[loss=0.3048, simple_loss=0.3727, pruned_loss=0.1185, over 8298.00 frames. ], tot_loss[loss=0.2891, simple_loss=0.3502, pruned_loss=0.114, over 1614899.80 frames. ], batch size: 23, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:43:44,280 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 01:43:50,924 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.02 vs. limit=5.0 +2023-02-06 01:44:07,432 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40116.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:44:15,425 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 3.016e+02 3.638e+02 4.428e+02 8.911e+02, threshold=7.276e+02, percent-clipped=8.0 +2023-02-06 01:44:16,233 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40129.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:44:18,939 INFO [train.py:901] (1/4) Epoch 5, batch 7800, loss[loss=0.3457, simple_loss=0.3855, pruned_loss=0.153, over 7108.00 frames. ], tot_loss[loss=0.2895, simple_loss=0.3504, pruned_loss=0.1143, over 1614543.27 frames. ], batch size: 71, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:44:50,150 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40177.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:44:54,263 INFO [train.py:901] (1/4) Epoch 5, batch 7850, loss[loss=0.2709, simple_loss=0.3442, pruned_loss=0.09883, over 8466.00 frames. ], tot_loss[loss=0.2882, simple_loss=0.3492, pruned_loss=0.1136, over 1610040.76 frames. ], batch size: 29, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:45:03,211 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40196.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:45:14,679 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:45:20,120 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40221.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:45:24,748 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.315e+02 3.285e+02 3.978e+02 4.753e+02 1.108e+03, threshold=7.955e+02, percent-clipped=4.0 +2023-02-06 01:45:28,283 INFO [train.py:901] (1/4) Epoch 5, batch 7900, loss[loss=0.3093, simple_loss=0.38, pruned_loss=0.1193, over 8644.00 frames. ], tot_loss[loss=0.2882, simple_loss=0.3491, pruned_loss=0.1137, over 1607296.72 frames. ], batch size: 39, lr: 1.42e-02, grad_scale: 16.0 +2023-02-06 01:45:30,494 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40236.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:45:35,936 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40244.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:46:02,028 INFO [train.py:901] (1/4) Epoch 5, batch 7950, loss[loss=0.2927, simple_loss=0.343, pruned_loss=0.1212, over 7928.00 frames. ], tot_loss[loss=0.286, simple_loss=0.3477, pruned_loss=0.1122, over 1608062.56 frames. ], batch size: 20, lr: 1.42e-02, grad_scale: 16.0 +2023-02-06 01:46:07,560 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40290.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:46:09,023 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40292.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:46:33,063 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 3.003e+02 3.931e+02 4.743e+02 9.937e+02, threshold=7.862e+02, percent-clipped=4.0 +2023-02-06 01:46:36,390 INFO [train.py:901] (1/4) Epoch 5, batch 8000, loss[loss=0.2773, simple_loss=0.3514, pruned_loss=0.1016, over 8326.00 frames. ], tot_loss[loss=0.2843, simple_loss=0.3468, pruned_loss=0.1108, over 1613735.60 frames. ], batch size: 25, lr: 1.42e-02, grad_scale: 16.0 +2023-02-06 01:46:46,580 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40348.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:47:03,147 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40372.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:47:10,472 INFO [train.py:901] (1/4) Epoch 5, batch 8050, loss[loss=0.1969, simple_loss=0.2619, pruned_loss=0.06595, over 7528.00 frames. ], tot_loss[loss=0.2824, simple_loss=0.3448, pruned_loss=0.11, over 1613654.45 frames. ], batch size: 18, lr: 1.42e-02, grad_scale: 8.0 +2023-02-06 01:47:20,239 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40397.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:47:43,883 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 01:47:48,096 INFO [train.py:901] (1/4) Epoch 6, batch 0, loss[loss=0.3063, simple_loss=0.3685, pruned_loss=0.1221, over 8334.00 frames. ], tot_loss[loss=0.3063, simple_loss=0.3685, pruned_loss=0.1221, over 8334.00 frames. ], batch size: 26, lr: 1.33e-02, grad_scale: 8.0 +2023-02-06 01:47:48,097 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 01:47:59,060 INFO [train.py:935] (1/4) Epoch 6, validation: loss=0.2203, simple_loss=0.3165, pruned_loss=0.06206, over 944034.00 frames. +2023-02-06 01:47:59,061 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6555MB +2023-02-06 01:48:07,792 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 3.052e+02 3.992e+02 5.098e+02 1.227e+03, threshold=7.983e+02, percent-clipped=7.0 +2023-02-06 01:48:12,971 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4260, 1.6380, 2.8931, 1.1442, 2.0539, 1.7634, 1.3566, 1.6793], + device='cuda:1'), covar=tensor([0.1684, 0.1970, 0.0651, 0.3465, 0.1469, 0.2533, 0.1702, 0.2075], + device='cuda:1'), in_proj_covar=tensor([0.0477, 0.0458, 0.0521, 0.0548, 0.0594, 0.0524, 0.0461, 0.0596], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 01:48:13,431 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 01:48:34,118 INFO [train.py:901] (1/4) Epoch 6, batch 50, loss[loss=0.2705, simple_loss=0.341, pruned_loss=0.09995, over 8200.00 frames. ], tot_loss[loss=0.2841, simple_loss=0.3476, pruned_loss=0.1103, over 365447.67 frames. ], batch size: 23, lr: 1.33e-02, grad_scale: 8.0 +2023-02-06 01:48:48,506 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 01:48:57,363 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40500.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:04,779 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:08,815 INFO [train.py:901] (1/4) Epoch 6, batch 100, loss[loss=0.247, simple_loss=0.3137, pruned_loss=0.09014, over 7699.00 frames. ], tot_loss[loss=0.2852, simple_loss=0.3484, pruned_loss=0.1111, over 642610.92 frames. ], batch size: 18, lr: 1.33e-02, grad_scale: 8.0 +2023-02-06 01:49:13,093 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 01:49:15,347 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40525.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:17,929 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.877e+02 3.627e+02 4.294e+02 7.601e+02, threshold=7.253e+02, percent-clipped=0.0 +2023-02-06 01:49:31,529 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40548.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:37,661 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40557.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:44,253 INFO [train.py:901] (1/4) Epoch 6, batch 150, loss[loss=0.2605, simple_loss=0.3393, pruned_loss=0.09085, over 8547.00 frames. ], tot_loss[loss=0.2862, simple_loss=0.3487, pruned_loss=0.1118, over 856977.01 frames. ], batch size: 31, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:49:49,721 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:54,329 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40580.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:50:19,249 INFO [train.py:901] (1/4) Epoch 6, batch 200, loss[loss=0.3075, simple_loss=0.3807, pruned_loss=0.1171, over 8510.00 frames. ], tot_loss[loss=0.2869, simple_loss=0.3498, pruned_loss=0.112, over 1029954.87 frames. ], batch size: 28, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:50:25,196 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 01:50:28,760 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 3.079e+02 3.898e+02 5.213e+02 9.157e+02, threshold=7.795e+02, percent-clipped=3.0 +2023-02-06 01:50:32,270 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:50:52,862 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8607, 1.9658, 2.1530, 1.7686, 1.2163, 2.3267, 0.4484, 1.2280], + device='cuda:1'), covar=tensor([0.3204, 0.1696, 0.0991, 0.2287, 0.5399, 0.0578, 0.5363, 0.2549], + device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0129, 0.0081, 0.0175, 0.0215, 0.0083, 0.0145, 0.0134], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:50:54,058 INFO [train.py:901] (1/4) Epoch 6, batch 250, loss[loss=0.2388, simple_loss=0.308, pruned_loss=0.08479, over 7699.00 frames. ], tot_loss[loss=0.2868, simple_loss=0.3492, pruned_loss=0.1122, over 1164529.09 frames. ], batch size: 18, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:50:56,282 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40669.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:50:58,372 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:51:04,072 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 01:51:12,281 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 01:51:13,018 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40692.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:51:15,104 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40695.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:51:29,232 INFO [train.py:901] (1/4) Epoch 6, batch 300, loss[loss=0.2225, simple_loss=0.2938, pruned_loss=0.07564, over 7532.00 frames. ], tot_loss[loss=0.2866, simple_loss=0.3493, pruned_loss=0.1119, over 1264590.72 frames. ], batch size: 18, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:51:38,587 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 3.025e+02 3.729e+02 4.724e+02 9.863e+02, threshold=7.458e+02, percent-clipped=3.0 +2023-02-06 01:51:47,273 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6824, 5.6357, 4.8353, 2.1056, 5.0045, 5.4384, 5.4263, 4.6764], + device='cuda:1'), covar=tensor([0.0575, 0.0427, 0.0839, 0.4547, 0.0572, 0.0424, 0.0882, 0.0546], + device='cuda:1'), in_proj_covar=tensor([0.0406, 0.0288, 0.0325, 0.0412, 0.0314, 0.0278, 0.0301, 0.0255], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 01:51:52,716 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40749.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:52:03,714 INFO [train.py:901] (1/4) Epoch 6, batch 350, loss[loss=0.288, simple_loss=0.3576, pruned_loss=0.1092, over 8106.00 frames. ], tot_loss[loss=0.2841, simple_loss=0.3472, pruned_loss=0.1105, over 1347446.51 frames. ], batch size: 23, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:52:32,430 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:52:38,237 INFO [train.py:901] (1/4) Epoch 6, batch 400, loss[loss=0.2974, simple_loss=0.3597, pruned_loss=0.1175, over 8483.00 frames. ], tot_loss[loss=0.286, simple_loss=0.3485, pruned_loss=0.1118, over 1405520.29 frames. ], batch size: 29, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:52:46,898 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.939e+02 3.080e+02 3.801e+02 5.022e+02 1.220e+03, threshold=7.601e+02, percent-clipped=4.0 +2023-02-06 01:53:03,002 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8936, 3.5722, 2.4649, 4.1734, 1.7545, 2.1517, 2.1518, 3.5281], + device='cuda:1'), covar=tensor([0.0799, 0.0747, 0.1196, 0.0213, 0.1661, 0.1718, 0.1681, 0.0864], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0246, 0.0281, 0.0223, 0.0247, 0.0278, 0.0280, 0.0256], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 01:53:04,907 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40854.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:53:12,738 INFO [train.py:901] (1/4) Epoch 6, batch 450, loss[loss=0.2015, simple_loss=0.2755, pruned_loss=0.06373, over 7537.00 frames. ], tot_loss[loss=0.2844, simple_loss=0.3474, pruned_loss=0.1106, over 1455082.82 frames. ], batch size: 18, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:53:24,458 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40883.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:53:47,844 INFO [train.py:901] (1/4) Epoch 6, batch 500, loss[loss=0.2628, simple_loss=0.3359, pruned_loss=0.09484, over 8330.00 frames. ], tot_loss[loss=0.2822, simple_loss=0.346, pruned_loss=0.1092, over 1490852.11 frames. ], batch size: 25, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:53:49,261 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40918.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:53:56,741 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:53:57,239 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 3.058e+02 3.738e+02 5.288e+02 8.550e+02, threshold=7.476e+02, percent-clipped=3.0 +2023-02-06 01:54:12,304 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:13,627 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40953.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:20,438 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6853, 1.9750, 1.5619, 2.2638, 1.3780, 1.4055, 1.6096, 1.9698], + device='cuda:1'), covar=tensor([0.0955, 0.0830, 0.1146, 0.0589, 0.1172, 0.1451, 0.0991, 0.0700], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0252, 0.0283, 0.0226, 0.0250, 0.0282, 0.0284, 0.0263], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 01:54:22,940 INFO [train.py:901] (1/4) Epoch 6, batch 550, loss[loss=0.2583, simple_loss=0.3344, pruned_loss=0.09108, over 8358.00 frames. ], tot_loss[loss=0.2829, simple_loss=0.3463, pruned_loss=0.1097, over 1517199.04 frames. ], batch size: 24, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:54:25,209 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:30,598 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:49,827 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:51,394 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 01:54:53,853 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9348, 2.1521, 1.6217, 2.7307, 1.1403, 1.3812, 1.6712, 2.2538], + device='cuda:1'), covar=tensor([0.0986, 0.1162, 0.1500, 0.0425, 0.1601, 0.1905, 0.1323, 0.1015], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0252, 0.0279, 0.0223, 0.0248, 0.0279, 0.0283, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 01:54:55,069 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41013.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:56,473 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41015.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:57,030 INFO [train.py:901] (1/4) Epoch 6, batch 600, loss[loss=0.2796, simple_loss=0.3494, pruned_loss=0.1049, over 8257.00 frames. ], tot_loss[loss=0.2791, simple_loss=0.3428, pruned_loss=0.1077, over 1532285.39 frames. ], batch size: 24, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:55:06,087 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.839e+02 3.515e+02 4.292e+02 8.268e+02, threshold=7.031e+02, percent-clipped=4.0 +2023-02-06 01:55:07,011 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41030.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:55:09,465 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 01:55:29,301 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41063.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:55:29,378 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41063.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:55:31,079 INFO [train.py:901] (1/4) Epoch 6, batch 650, loss[loss=0.288, simple_loss=0.3475, pruned_loss=0.1142, over 8027.00 frames. ], tot_loss[loss=0.278, simple_loss=0.3421, pruned_loss=0.107, over 1548803.23 frames. ], batch size: 22, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:55:46,208 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41088.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:55:59,433 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 01:56:05,813 INFO [train.py:901] (1/4) Epoch 6, batch 700, loss[loss=0.3367, simple_loss=0.3959, pruned_loss=0.1388, over 8541.00 frames. ], tot_loss[loss=0.2794, simple_loss=0.3435, pruned_loss=0.1077, over 1568679.34 frames. ], batch size: 28, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:56:08,624 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:56:14,237 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:56:14,737 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.998e+02 3.776e+02 4.654e+02 1.221e+03, threshold=7.553e+02, percent-clipped=4.0 +2023-02-06 01:56:40,070 INFO [train.py:901] (1/4) Epoch 6, batch 750, loss[loss=0.3778, simple_loss=0.4172, pruned_loss=0.1692, over 8620.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3434, pruned_loss=0.1079, over 1579065.73 frames. ], batch size: 34, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:56:42,121 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4681, 1.4364, 2.9820, 1.1802, 2.1275, 3.2641, 3.2606, 2.7230], + device='cuda:1'), covar=tensor([0.1273, 0.1560, 0.0429, 0.2287, 0.0828, 0.0306, 0.0428, 0.0686], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0266, 0.0228, 0.0262, 0.0234, 0.0207, 0.0242, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 01:56:52,792 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 01:57:00,953 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 01:57:15,161 INFO [train.py:901] (1/4) Epoch 6, batch 800, loss[loss=0.2682, simple_loss=0.3364, pruned_loss=0.1, over 8338.00 frames. ], tot_loss[loss=0.2804, simple_loss=0.3439, pruned_loss=0.1084, over 1591366.92 frames. ], batch size: 49, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:57:16,748 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2582, 1.7166, 1.6390, 0.6566, 1.6869, 1.2088, 0.2984, 1.5193], + device='cuda:1'), covar=tensor([0.0187, 0.0100, 0.0099, 0.0179, 0.0129, 0.0314, 0.0276, 0.0093], + device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0230, 0.0200, 0.0285, 0.0232, 0.0375, 0.0296, 0.0274], + device='cuda:1'), out_proj_covar=tensor([1.1118e-04, 7.6952e-05, 6.7294e-05, 9.6538e-05, 7.9966e-05, 1.3841e-04, + 1.0216e-04, 9.3394e-05], device='cuda:1') +2023-02-06 01:57:21,526 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41225.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:57:22,745 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41227.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:57:24,004 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.937e+02 3.578e+02 4.897e+02 8.076e+02, threshold=7.157e+02, percent-clipped=3.0 +2023-02-06 01:57:30,319 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6048, 2.1203, 2.1952, 0.9574, 2.1561, 1.3380, 0.6178, 1.7054], + device='cuda:1'), covar=tensor([0.0222, 0.0103, 0.0090, 0.0217, 0.0142, 0.0369, 0.0335, 0.0119], + device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0231, 0.0202, 0.0286, 0.0233, 0.0377, 0.0296, 0.0274], + device='cuda:1'), out_proj_covar=tensor([1.1151e-04, 7.6999e-05, 6.7681e-05, 9.6953e-05, 8.0298e-05, 1.3913e-04, + 1.0220e-04, 9.3277e-05], device='cuda:1') +2023-02-06 01:57:38,347 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:57:46,849 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41262.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:57:49,531 INFO [train.py:901] (1/4) Epoch 6, batch 850, loss[loss=0.2498, simple_loss=0.3238, pruned_loss=0.08785, over 8327.00 frames. ], tot_loss[loss=0.2808, simple_loss=0.3443, pruned_loss=0.1087, over 1599265.36 frames. ], batch size: 25, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:58:23,893 INFO [train.py:901] (1/4) Epoch 6, batch 900, loss[loss=0.2803, simple_loss=0.3424, pruned_loss=0.1091, over 7662.00 frames. ], tot_loss[loss=0.2807, simple_loss=0.3442, pruned_loss=0.1087, over 1600869.94 frames. ], batch size: 19, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:58:33,481 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.830e+02 2.871e+02 3.405e+02 4.321e+02 1.147e+03, threshold=6.810e+02, percent-clipped=2.0 +2023-02-06 01:58:42,516 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:58:53,862 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:58:58,558 INFO [train.py:901] (1/4) Epoch 6, batch 950, loss[loss=0.2992, simple_loss=0.3603, pruned_loss=0.119, over 8242.00 frames. ], tot_loss[loss=0.2786, simple_loss=0.3424, pruned_loss=0.1074, over 1600258.71 frames. ], batch size: 22, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:59:06,364 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41377.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:59:11,064 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41384.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:59:21,437 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 01:59:26,996 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41407.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:59:28,450 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41409.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:59:32,977 INFO [train.py:901] (1/4) Epoch 6, batch 1000, loss[loss=0.2638, simple_loss=0.336, pruned_loss=0.09581, over 8133.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3425, pruned_loss=0.1071, over 1606755.04 frames. ], batch size: 22, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:59:41,360 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.042e+02 3.293e+02 3.921e+02 5.074e+02 1.211e+03, threshold=7.843e+02, percent-clipped=6.0 +2023-02-06 01:59:55,295 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 02:00:06,251 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41464.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:00:07,521 INFO [train.py:901] (1/4) Epoch 6, batch 1050, loss[loss=0.2523, simple_loss=0.3173, pruned_loss=0.09367, over 7534.00 frames. ], tot_loss[loss=0.2788, simple_loss=0.3426, pruned_loss=0.1075, over 1603760.14 frames. ], batch size: 18, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:00:08,224 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 02:00:13,142 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41474.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:00:15,204 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4377, 1.2941, 1.4285, 1.2798, 0.8751, 1.3433, 1.1239, 1.0767], + device='cuda:1'), covar=tensor([0.0607, 0.1238, 0.1779, 0.1429, 0.0613, 0.1554, 0.0730, 0.0590], + device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0176, 0.0219, 0.0182, 0.0129, 0.0187, 0.0140, 0.0150], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 02:00:42,275 INFO [train.py:901] (1/4) Epoch 6, batch 1100, loss[loss=0.2786, simple_loss=0.3444, pruned_loss=0.1065, over 8482.00 frames. ], tot_loss[loss=0.2787, simple_loss=0.3428, pruned_loss=0.1073, over 1611339.42 frames. ], batch size: 25, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:00:46,694 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41522.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:00:51,106 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.799e+02 2.937e+02 3.488e+02 4.376e+02 9.981e+02, threshold=6.976e+02, percent-clipped=3.0 +2023-02-06 02:01:16,052 INFO [train.py:901] (1/4) Epoch 6, batch 1150, loss[loss=0.2084, simple_loss=0.2843, pruned_loss=0.06624, over 7781.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3412, pruned_loss=0.1062, over 1611518.93 frames. ], batch size: 19, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:01:18,802 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 02:01:25,448 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41579.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:01:28,764 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2916, 2.7048, 3.3222, 0.9358, 3.2333, 2.0564, 1.5221, 2.0740], + device='cuda:1'), covar=tensor([0.0256, 0.0125, 0.0086, 0.0278, 0.0157, 0.0289, 0.0362, 0.0161], + device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0235, 0.0204, 0.0280, 0.0228, 0.0374, 0.0294, 0.0273], + device='cuda:1'), out_proj_covar=tensor([1.0964e-04, 7.8724e-05, 6.8079e-05, 9.4752e-05, 7.8025e-05, 1.3747e-04, + 1.0116e-04, 9.2538e-05], device='cuda:1') +2023-02-06 02:01:32,278 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 02:01:33,975 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0830, 2.2909, 1.7696, 2.9772, 1.4520, 1.5308, 1.7955, 2.3468], + device='cuda:1'), covar=tensor([0.0789, 0.0911, 0.1355, 0.0389, 0.1320, 0.1691, 0.1341, 0.0884], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0249, 0.0280, 0.0224, 0.0244, 0.0276, 0.0283, 0.0256], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 02:01:38,083 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41598.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:01:50,379 INFO [train.py:901] (1/4) Epoch 6, batch 1200, loss[loss=0.2678, simple_loss=0.3281, pruned_loss=0.1037, over 7548.00 frames. ], tot_loss[loss=0.2778, simple_loss=0.3422, pruned_loss=0.1067, over 1617068.27 frames. ], batch size: 18, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:01:55,793 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:02:00,253 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 3.060e+02 3.864e+02 4.910e+02 1.275e+03, threshold=7.729e+02, percent-clipped=9.0 +2023-02-06 02:02:03,209 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41633.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:02:19,543 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41658.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:02:24,698 INFO [train.py:901] (1/4) Epoch 6, batch 1250, loss[loss=0.2144, simple_loss=0.2873, pruned_loss=0.07069, over 8098.00 frames. ], tot_loss[loss=0.2775, simple_loss=0.342, pruned_loss=0.1065, over 1621347.41 frames. ], batch size: 21, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:02:29,699 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:02:59,807 INFO [train.py:901] (1/4) Epoch 6, batch 1300, loss[loss=0.266, simple_loss=0.328, pruned_loss=0.1021, over 8107.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.3421, pruned_loss=0.1067, over 1616281.18 frames. ], batch size: 23, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:03:08,597 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 3.137e+02 4.028e+02 4.813e+02 9.668e+02, threshold=8.056e+02, percent-clipped=5.0 +2023-02-06 02:03:09,500 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41730.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:03:16,174 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.23 vs. limit=5.0 +2023-02-06 02:03:16,612 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0996, 2.0716, 1.9403, 1.8118, 1.3047, 1.8079, 2.0056, 1.9706], + device='cuda:1'), covar=tensor([0.0489, 0.1162, 0.1593, 0.1244, 0.0639, 0.1446, 0.0718, 0.0541], + device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0174, 0.0214, 0.0179, 0.0128, 0.0184, 0.0139, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 02:03:18,929 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 02:03:27,432 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:03:34,638 INFO [train.py:901] (1/4) Epoch 6, batch 1350, loss[loss=0.2634, simple_loss=0.3352, pruned_loss=0.09579, over 8249.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3419, pruned_loss=0.1071, over 1613070.33 frames. ], batch size: 24, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:03:40,202 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5042, 4.4891, 4.1030, 1.8269, 3.9147, 4.0091, 4.1809, 3.4419], + device='cuda:1'), covar=tensor([0.0819, 0.0661, 0.0927, 0.4792, 0.0734, 0.0678, 0.1405, 0.0963], + device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0290, 0.0319, 0.0405, 0.0310, 0.0279, 0.0301, 0.0255], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:03:42,952 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41778.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:04:00,419 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41803.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:04:08,478 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8978, 3.7843, 2.2800, 2.7611, 2.9093, 1.7787, 2.5354, 2.9143], + device='cuda:1'), covar=tensor([0.1327, 0.0299, 0.0865, 0.0668, 0.0659, 0.1170, 0.0914, 0.0844], + device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0237, 0.0310, 0.0294, 0.0311, 0.0311, 0.0335, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 02:04:09,616 INFO [train.py:901] (1/4) Epoch 6, batch 1400, loss[loss=0.2582, simple_loss=0.3303, pruned_loss=0.09307, over 8031.00 frames. ], tot_loss[loss=0.2775, simple_loss=0.3416, pruned_loss=0.1067, over 1616098.35 frames. ], batch size: 22, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:04:18,110 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.090e+02 3.079e+02 3.704e+02 4.589e+02 8.838e+02, threshold=7.407e+02, percent-clipped=2.0 +2023-02-06 02:04:22,366 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41835.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:04:39,965 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41860.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:04:44,581 INFO [train.py:901] (1/4) Epoch 6, batch 1450, loss[loss=0.268, simple_loss=0.3464, pruned_loss=0.0948, over 8349.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3408, pruned_loss=0.1064, over 1615397.51 frames. ], batch size: 24, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:04:47,870 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 02:05:02,018 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6099, 1.8693, 2.2907, 0.9724, 2.2975, 1.4660, 0.6385, 1.7025], + device='cuda:1'), covar=tensor([0.0208, 0.0132, 0.0100, 0.0203, 0.0139, 0.0345, 0.0284, 0.0119], + device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0232, 0.0202, 0.0282, 0.0229, 0.0372, 0.0294, 0.0269], + device='cuda:1'), out_proj_covar=tensor([1.0907e-04, 7.7481e-05, 6.7521e-05, 9.4855e-05, 7.8162e-05, 1.3630e-04, + 1.0143e-04, 9.1382e-05], device='cuda:1') +2023-02-06 02:05:18,593 INFO [train.py:901] (1/4) Epoch 6, batch 1500, loss[loss=0.269, simple_loss=0.3313, pruned_loss=0.1034, over 7406.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.3419, pruned_loss=0.1072, over 1615565.81 frames. ], batch size: 17, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:05:24,650 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41924.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:05:27,834 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.922e+02 3.542e+02 4.432e+02 1.007e+03, threshold=7.084e+02, percent-clipped=2.0 +2023-02-06 02:05:39,393 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8346, 2.1890, 1.6813, 2.7655, 1.2330, 1.4125, 1.5479, 2.1555], + device='cuda:1'), covar=tensor([0.1040, 0.1037, 0.1372, 0.0446, 0.1596, 0.1998, 0.1443, 0.1002], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0247, 0.0277, 0.0223, 0.0245, 0.0276, 0.0278, 0.0255], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 02:05:53,225 INFO [train.py:901] (1/4) Epoch 6, batch 1550, loss[loss=0.2771, simple_loss=0.3399, pruned_loss=0.1072, over 8075.00 frames. ], tot_loss[loss=0.2785, simple_loss=0.3421, pruned_loss=0.1074, over 1615942.13 frames. ], batch size: 21, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:05:53,368 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41966.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:06:28,445 INFO [train.py:901] (1/4) Epoch 6, batch 1600, loss[loss=0.3102, simple_loss=0.3828, pruned_loss=0.1188, over 8468.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3419, pruned_loss=0.1071, over 1615151.77 frames. ], batch size: 25, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:06:28,510 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42016.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:06:37,881 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.135e+02 3.132e+02 3.836e+02 5.392e+02 3.005e+03, threshold=7.672e+02, percent-clipped=11.0 +2023-02-06 02:06:42,325 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.00 vs. limit=5.0 +2023-02-06 02:07:03,792 INFO [train.py:901] (1/4) Epoch 6, batch 1650, loss[loss=0.2943, simple_loss=0.3511, pruned_loss=0.1188, over 8457.00 frames. ], tot_loss[loss=0.2763, simple_loss=0.3409, pruned_loss=0.1058, over 1621178.61 frames. ], batch size: 27, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:07:39,088 INFO [train.py:901] (1/4) Epoch 6, batch 1700, loss[loss=0.3106, simple_loss=0.3729, pruned_loss=0.1242, over 7933.00 frames. ], tot_loss[loss=0.2756, simple_loss=0.3401, pruned_loss=0.1055, over 1615724.01 frames. ], batch size: 20, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:07:47,888 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 2.826e+02 3.670e+02 4.452e+02 1.049e+03, threshold=7.339e+02, percent-clipped=2.0 +2023-02-06 02:07:49,313 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42131.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:07:52,041 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0193, 1.3209, 4.2289, 1.4987, 3.6804, 3.4891, 3.8093, 3.6912], + device='cuda:1'), covar=tensor([0.0485, 0.3759, 0.0453, 0.2822, 0.1173, 0.0706, 0.0502, 0.0588], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0504, 0.0442, 0.0435, 0.0500, 0.0414, 0.0412, 0.0463], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 02:08:05,249 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4955, 1.5421, 1.5745, 1.4151, 0.9402, 1.7287, 0.1018, 0.9204], + device='cuda:1'), covar=tensor([0.2838, 0.2055, 0.0843, 0.1787, 0.5701, 0.0754, 0.4306, 0.2363], + device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0132, 0.0082, 0.0179, 0.0216, 0.0083, 0.0145, 0.0135], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:08:14,479 INFO [train.py:901] (1/4) Epoch 6, batch 1750, loss[loss=0.2085, simple_loss=0.289, pruned_loss=0.06405, over 6383.00 frames. ], tot_loss[loss=0.2775, simple_loss=0.3412, pruned_loss=0.107, over 1613125.48 frames. ], batch size: 14, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:08:49,383 INFO [train.py:901] (1/4) Epoch 6, batch 1800, loss[loss=0.298, simple_loss=0.3571, pruned_loss=0.1195, over 8135.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.3409, pruned_loss=0.1061, over 1613036.20 frames. ], batch size: 22, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:08:59,173 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 3.032e+02 3.540e+02 4.353e+02 2.015e+03, threshold=7.080e+02, percent-clipped=5.0 +2023-02-06 02:09:24,927 INFO [train.py:901] (1/4) Epoch 6, batch 1850, loss[loss=0.2923, simple_loss=0.3629, pruned_loss=0.1109, over 8334.00 frames. ], tot_loss[loss=0.2778, simple_loss=0.3411, pruned_loss=0.1072, over 1608781.78 frames. ], batch size: 49, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:09:26,411 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42268.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:09:55,347 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42310.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:09:59,286 INFO [train.py:901] (1/4) Epoch 6, batch 1900, loss[loss=0.2943, simple_loss=0.3566, pruned_loss=0.1161, over 8189.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.3405, pruned_loss=0.1064, over 1610222.21 frames. ], batch size: 23, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:10:08,776 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.992e+02 2.715e+02 3.297e+02 4.142e+02 7.213e+02, threshold=6.594e+02, percent-clipped=2.0 +2023-02-06 02:10:23,963 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 02:10:34,055 INFO [train.py:901] (1/4) Epoch 6, batch 1950, loss[loss=0.2384, simple_loss=0.3106, pruned_loss=0.08315, over 8447.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3386, pruned_loss=0.1048, over 1614011.66 frames. ], batch size: 27, lr: 1.30e-02, grad_scale: 16.0 +2023-02-06 02:10:36,639 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 02:10:46,215 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42383.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:10:48,986 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42387.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:10:56,196 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 02:10:56,323 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42397.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:11:06,414 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42412.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:11:08,983 INFO [train.py:901] (1/4) Epoch 6, batch 2000, loss[loss=0.2732, simple_loss=0.3493, pruned_loss=0.09859, over 8499.00 frames. ], tot_loss[loss=0.273, simple_loss=0.3378, pruned_loss=0.1041, over 1615000.12 frames. ], batch size: 28, lr: 1.30e-02, grad_scale: 16.0 +2023-02-06 02:11:15,185 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42425.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:11:18,300 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.766e+02 3.581e+02 4.303e+02 8.011e+02, threshold=7.162e+02, percent-clipped=3.0 +2023-02-06 02:11:19,861 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2184, 1.4915, 1.2475, 1.9565, 0.7764, 1.0720, 1.1202, 1.5071], + device='cuda:1'), covar=tensor([0.1208, 0.0984, 0.1568, 0.0605, 0.1527, 0.2052, 0.1230, 0.0878], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0251, 0.0277, 0.0222, 0.0245, 0.0277, 0.0279, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 02:11:21,925 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4526, 1.7690, 1.8603, 0.8554, 1.9726, 1.3433, 0.4818, 1.6822], + device='cuda:1'), covar=tensor([0.0194, 0.0128, 0.0127, 0.0228, 0.0133, 0.0403, 0.0332, 0.0112], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0242, 0.0210, 0.0294, 0.0237, 0.0386, 0.0308, 0.0280], + device='cuda:1'), out_proj_covar=tensor([1.1273e-04, 8.0154e-05, 7.0017e-05, 9.8503e-05, 8.0558e-05, 1.4074e-04, + 1.0550e-04, 9.4484e-05], device='cuda:1') +2023-02-06 02:11:43,879 INFO [train.py:901] (1/4) Epoch 6, batch 2050, loss[loss=0.243, simple_loss=0.3181, pruned_loss=0.08392, over 8353.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3391, pruned_loss=0.1052, over 1612380.34 frames. ], batch size: 24, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:11:52,269 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-02-06 02:12:17,690 INFO [train.py:901] (1/4) Epoch 6, batch 2100, loss[loss=0.2857, simple_loss=0.3503, pruned_loss=0.1106, over 8359.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3395, pruned_loss=0.1052, over 1615646.65 frames. ], batch size: 24, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:12:23,854 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:12:24,530 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5626, 4.3475, 3.9875, 1.7089, 3.9743, 3.9785, 4.1944, 3.6379], + device='cuda:1'), covar=tensor([0.0736, 0.0550, 0.0948, 0.4574, 0.0694, 0.0804, 0.0998, 0.0893], + device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0293, 0.0324, 0.0402, 0.0316, 0.0283, 0.0306, 0.0258], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:12:27,686 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.916e+02 3.481e+02 4.572e+02 1.310e+03, threshold=6.962e+02, percent-clipped=2.0 +2023-02-06 02:12:50,563 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1787, 1.0637, 1.0661, 1.1838, 0.8014, 1.2485, 0.1644, 0.9399], + device='cuda:1'), covar=tensor([0.2830, 0.2278, 0.1212, 0.1919, 0.5633, 0.0940, 0.4437, 0.2122], + device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0133, 0.0085, 0.0180, 0.0221, 0.0084, 0.0142, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:12:52,386 INFO [train.py:901] (1/4) Epoch 6, batch 2150, loss[loss=0.281, simple_loss=0.3519, pruned_loss=0.105, over 8563.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3399, pruned_loss=0.1056, over 1615046.09 frames. ], batch size: 49, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:12:58,003 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0190, 1.5627, 2.3544, 1.9490, 2.0195, 1.8451, 1.4727, 0.6107], + device='cuda:1'), covar=tensor([0.1966, 0.2149, 0.0538, 0.1069, 0.0831, 0.1146, 0.1083, 0.2000], + device='cuda:1'), in_proj_covar=tensor([0.0777, 0.0715, 0.0619, 0.0714, 0.0803, 0.0661, 0.0623, 0.0655], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:13:27,037 INFO [train.py:901] (1/4) Epoch 6, batch 2200, loss[loss=0.3055, simple_loss=0.3705, pruned_loss=0.1202, over 8245.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.3404, pruned_loss=0.1064, over 1612463.88 frames. ], batch size: 24, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:13:36,152 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.877e+02 3.020e+02 3.729e+02 5.072e+02 1.122e+03, threshold=7.459e+02, percent-clipped=5.0 +2023-02-06 02:13:43,115 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42639.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:13:45,254 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-06 02:13:59,701 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42664.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:14:00,834 INFO [train.py:901] (1/4) Epoch 6, batch 2250, loss[loss=0.2403, simple_loss=0.3213, pruned_loss=0.07969, over 8471.00 frames. ], tot_loss[loss=0.2763, simple_loss=0.3405, pruned_loss=0.106, over 1613612.68 frames. ], batch size: 25, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:14:01,614 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:14:06,424 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8561, 1.5491, 3.2448, 1.3179, 2.1961, 3.6225, 3.5443, 3.0533], + device='cuda:1'), covar=tensor([0.0972, 0.1283, 0.0320, 0.1803, 0.0720, 0.0200, 0.0295, 0.0531], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0268, 0.0227, 0.0263, 0.0234, 0.0207, 0.0249, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 02:14:11,438 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 02:14:11,882 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42681.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:14:26,793 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 02:14:29,250 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42706.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:14:35,836 INFO [train.py:901] (1/4) Epoch 6, batch 2300, loss[loss=0.2706, simple_loss=0.3215, pruned_loss=0.1099, over 7798.00 frames. ], tot_loss[loss=0.2773, simple_loss=0.3411, pruned_loss=0.1067, over 1615868.85 frames. ], batch size: 19, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:14:45,247 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.977e+02 3.532e+02 4.435e+02 7.362e+02, threshold=7.063e+02, percent-clipped=0.0 +2023-02-06 02:14:53,243 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42741.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:15:11,259 INFO [train.py:901] (1/4) Epoch 6, batch 2350, loss[loss=0.3934, simple_loss=0.4259, pruned_loss=0.1805, over 6611.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.3403, pruned_loss=0.1063, over 1610967.21 frames. ], batch size: 71, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:15:46,903 INFO [train.py:901] (1/4) Epoch 6, batch 2400, loss[loss=0.2851, simple_loss=0.3415, pruned_loss=0.1144, over 7784.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.3404, pruned_loss=0.1063, over 1610348.63 frames. ], batch size: 19, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:15:56,300 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.044e+02 3.099e+02 3.712e+02 4.452e+02 1.076e+03, threshold=7.425e+02, percent-clipped=4.0 +2023-02-06 02:16:14,313 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:16:20,855 INFO [train.py:901] (1/4) Epoch 6, batch 2450, loss[loss=0.2986, simple_loss=0.3658, pruned_loss=0.1157, over 8608.00 frames. ], tot_loss[loss=0.277, simple_loss=0.3411, pruned_loss=0.1064, over 1613213.79 frames. ], batch size: 39, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:16:22,281 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:16:29,141 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42877.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:16:42,215 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42897.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:16:47,007 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9758, 1.8425, 3.2137, 2.6440, 2.5482, 1.7896, 1.4033, 1.5131], + device='cuda:1'), covar=tensor([0.3124, 0.3116, 0.0619, 0.1358, 0.1492, 0.1681, 0.1643, 0.2564], + device='cuda:1'), in_proj_covar=tensor([0.0788, 0.0721, 0.0620, 0.0720, 0.0806, 0.0664, 0.0628, 0.0656], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:16:54,606 INFO [train.py:901] (1/4) Epoch 6, batch 2500, loss[loss=0.2524, simple_loss=0.3311, pruned_loss=0.08685, over 8104.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.3405, pruned_loss=0.1062, over 1613972.75 frames. ], batch size: 23, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:17:05,200 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 3.094e+02 4.004e+02 4.995e+02 1.056e+03, threshold=8.009e+02, percent-clipped=4.0 +2023-02-06 02:17:15,537 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4742, 1.5056, 1.6598, 1.3138, 0.9396, 1.7329, 0.0979, 0.9137], + device='cuda:1'), covar=tensor([0.2693, 0.2294, 0.0908, 0.2120, 0.5659, 0.0714, 0.4251, 0.2678], + device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0136, 0.0085, 0.0185, 0.0223, 0.0086, 0.0144, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:17:15,681 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 02:17:25,672 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 02:17:29,433 INFO [train.py:901] (1/4) Epoch 6, batch 2550, loss[loss=0.2688, simple_loss=0.3448, pruned_loss=0.09637, over 8512.00 frames. ], tot_loss[loss=0.2771, simple_loss=0.3418, pruned_loss=0.1063, over 1617448.43 frames. ], batch size: 28, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:17:41,643 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42983.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:18:01,216 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43011.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:18:04,437 INFO [train.py:901] (1/4) Epoch 6, batch 2600, loss[loss=0.276, simple_loss=0.3505, pruned_loss=0.1007, over 8445.00 frames. ], tot_loss[loss=0.2795, simple_loss=0.3436, pruned_loss=0.1077, over 1618972.94 frames. ], batch size: 27, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:18:13,988 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 3.052e+02 3.779e+02 5.019e+02 1.784e+03, threshold=7.558e+02, percent-clipped=4.0 +2023-02-06 02:18:39,587 INFO [train.py:901] (1/4) Epoch 6, batch 2650, loss[loss=0.2611, simple_loss=0.3188, pruned_loss=0.1017, over 7212.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3432, pruned_loss=0.1067, over 1621764.92 frames. ], batch size: 16, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:18:47,154 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7165, 2.4962, 2.9271, 2.0103, 1.2711, 2.7986, 0.7220, 1.6378], + device='cuda:1'), covar=tensor([0.2862, 0.2271, 0.0798, 0.3144, 0.6429, 0.0559, 0.5718, 0.2694], + device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0131, 0.0081, 0.0179, 0.0217, 0.0082, 0.0141, 0.0133], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:19:05,111 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1280, 1.0844, 1.1230, 1.0736, 0.7533, 1.2249, 0.0337, 0.7574], + device='cuda:1'), covar=tensor([0.2809, 0.2520, 0.1018, 0.1884, 0.5370, 0.0792, 0.4713, 0.2672], + device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0133, 0.0083, 0.0182, 0.0219, 0.0084, 0.0144, 0.0136], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:19:11,158 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43112.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:19:14,342 INFO [train.py:901] (1/4) Epoch 6, batch 2700, loss[loss=0.2548, simple_loss=0.3144, pruned_loss=0.09761, over 7302.00 frames. ], tot_loss[loss=0.2799, simple_loss=0.3444, pruned_loss=0.1078, over 1622036.20 frames. ], batch size: 16, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:19:15,202 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4343, 1.7855, 2.8314, 1.1531, 2.0602, 1.7283, 1.4900, 1.7450], + device='cuda:1'), covar=tensor([0.1573, 0.1954, 0.0679, 0.3606, 0.1395, 0.2556, 0.1590, 0.2114], + device='cuda:1'), in_proj_covar=tensor([0.0472, 0.0462, 0.0529, 0.0542, 0.0594, 0.0528, 0.0449, 0.0601], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 02:19:20,969 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43126.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:19:23,445 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.935e+02 3.532e+02 4.548e+02 1.003e+03, threshold=7.064e+02, percent-clipped=2.0 +2023-02-06 02:19:28,350 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43137.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:19:45,814 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-06 02:19:47,954 INFO [train.py:901] (1/4) Epoch 6, batch 2750, loss[loss=0.2918, simple_loss=0.367, pruned_loss=0.1084, over 8027.00 frames. ], tot_loss[loss=0.2779, simple_loss=0.3427, pruned_loss=0.1065, over 1616214.61 frames. ], batch size: 22, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:19:57,321 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5008, 1.9396, 2.1468, 1.0909, 2.2752, 1.4287, 0.8073, 1.7593], + device='cuda:1'), covar=tensor([0.0258, 0.0110, 0.0071, 0.0202, 0.0123, 0.0351, 0.0288, 0.0125], + device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0235, 0.0206, 0.0288, 0.0227, 0.0381, 0.0298, 0.0273], + device='cuda:1'), out_proj_covar=tensor([1.0839e-04, 7.7382e-05, 6.8384e-05, 9.5718e-05, 7.6279e-05, 1.3795e-04, + 1.0173e-04, 9.1606e-05], device='cuda:1') +2023-02-06 02:20:02,613 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3123, 1.9539, 3.2599, 1.0262, 2.6964, 1.6066, 1.6270, 2.0995], + device='cuda:1'), covar=tensor([0.1911, 0.1958, 0.0811, 0.3787, 0.1420, 0.2912, 0.1770, 0.2452], + device='cuda:1'), in_proj_covar=tensor([0.0469, 0.0460, 0.0526, 0.0535, 0.0589, 0.0523, 0.0444, 0.0590], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 02:20:22,671 INFO [train.py:901] (1/4) Epoch 6, batch 2800, loss[loss=0.2943, simple_loss=0.3663, pruned_loss=0.1111, over 8480.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3429, pruned_loss=0.1066, over 1616163.94 frames. ], batch size: 28, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:20:26,179 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43221.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:20:32,060 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.696e+02 3.315e+02 4.271e+02 8.534e+02, threshold=6.630e+02, percent-clipped=4.0 +2023-02-06 02:20:39,132 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43239.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:20:40,379 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43241.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:20:55,849 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43264.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:20:56,925 INFO [train.py:901] (1/4) Epoch 6, batch 2850, loss[loss=0.2828, simple_loss=0.3531, pruned_loss=0.1063, over 8359.00 frames. ], tot_loss[loss=0.2752, simple_loss=0.3409, pruned_loss=0.1047, over 1615947.91 frames. ], batch size: 24, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:21:11,145 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6376, 4.0352, 2.1910, 1.9841, 2.6694, 1.4683, 2.1388, 2.8587], + device='cuda:1'), covar=tensor([0.1713, 0.0346, 0.1083, 0.1008, 0.0831, 0.1603, 0.1420, 0.0844], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0237, 0.0312, 0.0304, 0.0317, 0.0313, 0.0340, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 02:21:31,689 INFO [train.py:901] (1/4) Epoch 6, batch 2900, loss[loss=0.2266, simple_loss=0.2993, pruned_loss=0.07702, over 8233.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3401, pruned_loss=0.1049, over 1613052.90 frames. ], batch size: 22, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:21:41,574 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.996e+02 3.885e+02 4.976e+02 9.964e+02, threshold=7.771e+02, percent-clipped=9.0 +2023-02-06 02:21:46,023 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:22:00,469 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43356.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:22:01,693 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 02:22:07,135 INFO [train.py:901] (1/4) Epoch 6, batch 2950, loss[loss=0.2612, simple_loss=0.3489, pruned_loss=0.08676, over 8316.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3417, pruned_loss=0.106, over 1616121.47 frames. ], batch size: 25, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:22:17,904 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43382.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:22:35,335 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43407.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:22:41,761 INFO [train.py:901] (1/4) Epoch 6, batch 3000, loss[loss=0.3135, simple_loss=0.3755, pruned_loss=0.1258, over 8564.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3413, pruned_loss=0.1058, over 1620685.57 frames. ], batch size: 34, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:22:41,762 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 02:22:53,878 INFO [train.py:935] (1/4) Epoch 6, validation: loss=0.2158, simple_loss=0.3124, pruned_loss=0.05962, over 944034.00 frames. +2023-02-06 02:22:53,879 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6607MB +2023-02-06 02:23:03,877 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 3.288e+02 4.080e+02 5.339e+02 1.082e+03, threshold=8.161e+02, percent-clipped=5.0 +2023-02-06 02:23:20,244 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4297, 1.9259, 2.1153, 0.9446, 2.2492, 1.3988, 0.4799, 1.7287], + device='cuda:1'), covar=tensor([0.0251, 0.0111, 0.0110, 0.0198, 0.0116, 0.0333, 0.0322, 0.0110], + device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0238, 0.0208, 0.0293, 0.0231, 0.0384, 0.0300, 0.0274], + device='cuda:1'), out_proj_covar=tensor([1.0977e-04, 7.8215e-05, 6.8408e-05, 9.7167e-05, 7.7805e-05, 1.3890e-04, + 1.0211e-04, 9.1620e-05], device='cuda:1') +2023-02-06 02:23:28,759 INFO [train.py:901] (1/4) Epoch 6, batch 3050, loss[loss=0.2896, simple_loss=0.3508, pruned_loss=0.1142, over 8677.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3411, pruned_loss=0.1059, over 1623149.18 frames. ], batch size: 34, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:24:03,332 INFO [train.py:901] (1/4) Epoch 6, batch 3100, loss[loss=0.2802, simple_loss=0.3323, pruned_loss=0.114, over 7980.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3404, pruned_loss=0.1056, over 1619870.11 frames. ], batch size: 21, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:24:12,759 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.871e+02 3.509e+02 4.582e+02 1.148e+03, threshold=7.017e+02, percent-clipped=4.0 +2023-02-06 02:24:14,186 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:24:36,351 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7482, 1.3143, 3.3479, 1.2062, 2.1382, 3.7106, 3.7033, 3.1305], + device='cuda:1'), covar=tensor([0.1143, 0.1587, 0.0352, 0.1989, 0.0896, 0.0221, 0.0392, 0.0653], + device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0274, 0.0228, 0.0265, 0.0244, 0.0212, 0.0252, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 02:24:38,268 INFO [train.py:901] (1/4) Epoch 6, batch 3150, loss[loss=0.3752, simple_loss=0.4011, pruned_loss=0.1747, over 6597.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3409, pruned_loss=0.1057, over 1615845.67 frames. ], batch size: 73, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:24:57,078 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:25:10,983 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43612.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:25:13,476 INFO [train.py:901] (1/4) Epoch 6, batch 3200, loss[loss=0.3202, simple_loss=0.3762, pruned_loss=0.1321, over 8328.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3412, pruned_loss=0.1056, over 1619157.02 frames. ], batch size: 26, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:25:14,376 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43617.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:25:23,589 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.828e+02 3.409e+02 4.222e+02 1.719e+03, threshold=6.818e+02, percent-clipped=4.0 +2023-02-06 02:25:28,578 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43637.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:25:49,155 INFO [train.py:901] (1/4) Epoch 6, batch 3250, loss[loss=0.4689, simple_loss=0.4618, pruned_loss=0.238, over 6916.00 frames. ], tot_loss[loss=0.2767, simple_loss=0.3418, pruned_loss=0.1058, over 1621802.32 frames. ], batch size: 75, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:26:23,420 INFO [train.py:901] (1/4) Epoch 6, batch 3300, loss[loss=0.2374, simple_loss=0.3033, pruned_loss=0.08579, over 7971.00 frames. ], tot_loss[loss=0.2751, simple_loss=0.34, pruned_loss=0.1051, over 1614566.89 frames. ], batch size: 21, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:26:33,006 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.968e+02 3.670e+02 5.054e+02 9.057e+02, threshold=7.341e+02, percent-clipped=6.0 +2023-02-06 02:26:57,497 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1651, 3.0811, 2.8810, 1.3841, 2.7913, 2.8503, 2.9413, 2.5807], + device='cuda:1'), covar=tensor([0.1159, 0.0799, 0.1124, 0.4728, 0.1024, 0.1042, 0.1468, 0.1131], + device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0296, 0.0333, 0.0412, 0.0321, 0.0287, 0.0318, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:26:58,037 INFO [train.py:901] (1/4) Epoch 6, batch 3350, loss[loss=0.3038, simple_loss=0.3623, pruned_loss=0.1226, over 8242.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3414, pruned_loss=0.1057, over 1615990.21 frames. ], batch size: 24, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:27:25,489 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43805.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:27:33,326 INFO [train.py:901] (1/4) Epoch 6, batch 3400, loss[loss=0.3301, simple_loss=0.3946, pruned_loss=0.1328, over 8348.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3415, pruned_loss=0.1057, over 1613549.18 frames. ], batch size: 26, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:27:42,440 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.693e+02 3.397e+02 4.441e+02 9.371e+02, threshold=6.793e+02, percent-clipped=2.0 +2023-02-06 02:28:07,540 INFO [train.py:901] (1/4) Epoch 6, batch 3450, loss[loss=0.2739, simple_loss=0.3245, pruned_loss=0.1116, over 7820.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3414, pruned_loss=0.1056, over 1614208.45 frames. ], batch size: 20, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:28:14,344 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:28:42,237 INFO [train.py:901] (1/4) Epoch 6, batch 3500, loss[loss=0.2029, simple_loss=0.2865, pruned_loss=0.05965, over 7981.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.34, pruned_loss=0.1047, over 1615385.32 frames. ], batch size: 21, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:28:50,512 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:28:52,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 3.111e+02 3.775e+02 4.956e+02 7.195e+02, threshold=7.550e+02, percent-clipped=1.0 +2023-02-06 02:28:59,190 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 02:29:06,064 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 02:29:16,408 INFO [train.py:901] (1/4) Epoch 6, batch 3550, loss[loss=0.2747, simple_loss=0.3352, pruned_loss=0.1071, over 8077.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.3399, pruned_loss=0.1055, over 1612681.18 frames. ], batch size: 21, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:29:24,736 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43977.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:29:34,248 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43991.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:29:52,633 INFO [train.py:901] (1/4) Epoch 6, batch 3600, loss[loss=0.2431, simple_loss=0.2953, pruned_loss=0.09541, over 7812.00 frames. ], tot_loss[loss=0.2743, simple_loss=0.3388, pruned_loss=0.1049, over 1608279.51 frames. ], batch size: 19, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:30:02,263 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 2.983e+02 3.632e+02 4.470e+02 1.452e+03, threshold=7.265e+02, percent-clipped=1.0 +2023-02-06 02:30:27,003 INFO [train.py:901] (1/4) Epoch 6, batch 3650, loss[loss=0.2804, simple_loss=0.3406, pruned_loss=0.1101, over 7532.00 frames. ], tot_loss[loss=0.2735, simple_loss=0.3384, pruned_loss=0.1043, over 1606367.52 frames. ], batch size: 18, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:31:00,588 INFO [train.py:901] (1/4) Epoch 6, batch 3700, loss[loss=0.3117, simple_loss=0.3692, pruned_loss=0.1272, over 8323.00 frames. ], tot_loss[loss=0.2751, simple_loss=0.3397, pruned_loss=0.1053, over 1611353.84 frames. ], batch size: 49, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:31:01,285 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 02:31:11,218 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.832e+02 3.019e+02 3.651e+02 4.413e+02 8.839e+02, threshold=7.303e+02, percent-clipped=3.0 +2023-02-06 02:31:23,984 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44149.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:31:35,822 INFO [train.py:901] (1/4) Epoch 6, batch 3750, loss[loss=0.2792, simple_loss=0.344, pruned_loss=0.1072, over 8354.00 frames. ], tot_loss[loss=0.2744, simple_loss=0.3389, pruned_loss=0.1049, over 1608153.28 frames. ], batch size: 24, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:31:49,949 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8377, 1.6116, 3.1622, 1.2690, 2.1795, 3.3533, 3.4387, 2.6450], + device='cuda:1'), covar=tensor([0.1106, 0.1401, 0.0402, 0.2082, 0.0828, 0.0426, 0.0487, 0.0911], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0272, 0.0225, 0.0267, 0.0237, 0.0211, 0.0248, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 02:32:09,332 INFO [train.py:901] (1/4) Epoch 6, batch 3800, loss[loss=0.3029, simple_loss=0.3589, pruned_loss=0.1234, over 8679.00 frames. ], tot_loss[loss=0.2744, simple_loss=0.3385, pruned_loss=0.1051, over 1606609.08 frames. ], batch size: 34, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:32:19,589 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 3.008e+02 3.761e+02 4.930e+02 1.044e+03, threshold=7.521e+02, percent-clipped=7.0 +2023-02-06 02:32:32,456 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:32:44,266 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44264.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:32:45,463 INFO [train.py:901] (1/4) Epoch 6, batch 3850, loss[loss=0.254, simple_loss=0.318, pruned_loss=0.09497, over 7794.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.3392, pruned_loss=0.1057, over 1607941.77 frames. ], batch size: 19, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:32:47,013 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4422, 1.3620, 4.6399, 1.7410, 4.0132, 3.8612, 4.1696, 4.0394], + device='cuda:1'), covar=tensor([0.0451, 0.3879, 0.0434, 0.2715, 0.1093, 0.0710, 0.0443, 0.0573], + device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0504, 0.0449, 0.0440, 0.0506, 0.0416, 0.0423, 0.0478], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 02:32:48,960 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44271.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:32:49,789 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44272.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:32:51,204 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4419, 1.6880, 1.5503, 1.3475, 1.2157, 1.4070, 1.8687, 1.4599], + device='cuda:1'), covar=tensor([0.0553, 0.1172, 0.1812, 0.1396, 0.0618, 0.1518, 0.0691, 0.0604], + device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0171, 0.0213, 0.0177, 0.0121, 0.0181, 0.0137, 0.0148], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 02:33:02,827 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 02:33:20,476 INFO [train.py:901] (1/4) Epoch 6, batch 3900, loss[loss=0.2864, simple_loss=0.3562, pruned_loss=0.1083, over 8588.00 frames. ], tot_loss[loss=0.2757, simple_loss=0.3399, pruned_loss=0.1057, over 1611025.83 frames. ], batch size: 31, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:33:23,926 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44321.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:33:30,569 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.224e+02 2.909e+02 3.535e+02 4.398e+02 8.405e+02, threshold=7.069e+02, percent-clipped=2.0 +2023-02-06 02:33:56,224 INFO [train.py:901] (1/4) Epoch 6, batch 3950, loss[loss=0.3221, simple_loss=0.3767, pruned_loss=0.1337, over 8523.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3402, pruned_loss=0.1057, over 1617549.18 frames. ], batch size: 28, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:34:09,754 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44386.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:34:17,818 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1665, 1.3271, 2.3328, 1.0664, 2.1149, 2.5353, 2.4881, 2.1422], + device='cuda:1'), covar=tensor([0.0977, 0.1034, 0.0475, 0.1835, 0.0531, 0.0359, 0.0507, 0.0725], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0272, 0.0226, 0.0267, 0.0235, 0.0211, 0.0248, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 02:34:30,798 INFO [train.py:901] (1/4) Epoch 6, batch 4000, loss[loss=0.2603, simple_loss=0.3404, pruned_loss=0.09009, over 8194.00 frames. ], tot_loss[loss=0.2742, simple_loss=0.3389, pruned_loss=0.1047, over 1616737.19 frames. ], batch size: 23, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:34:40,322 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.014e+02 2.805e+02 3.702e+02 4.857e+02 8.487e+02, threshold=7.405e+02, percent-clipped=7.0 +2023-02-06 02:34:44,616 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44436.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:34:59,497 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.42 vs. limit=5.0 +2023-02-06 02:35:05,731 INFO [train.py:901] (1/4) Epoch 6, batch 4050, loss[loss=0.2695, simple_loss=0.3378, pruned_loss=0.1006, over 8086.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3379, pruned_loss=0.1037, over 1613457.39 frames. ], batch size: 21, lr: 1.27e-02, grad_scale: 16.0 +2023-02-06 02:35:39,370 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3274, 2.2447, 1.6805, 2.0913, 1.8135, 1.4065, 1.6672, 1.8393], + device='cuda:1'), covar=tensor([0.0961, 0.0325, 0.0811, 0.0358, 0.0516, 0.1026, 0.0642, 0.0578], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0236, 0.0310, 0.0306, 0.0322, 0.0314, 0.0337, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 02:35:41,172 INFO [train.py:901] (1/4) Epoch 6, batch 4100, loss[loss=0.2855, simple_loss=0.3536, pruned_loss=0.1087, over 8282.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3383, pruned_loss=0.1036, over 1614914.30 frames. ], batch size: 23, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:35:44,018 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:35:50,475 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 3.131e+02 3.987e+02 5.314e+02 1.327e+03, threshold=7.973e+02, percent-clipped=4.0 +2023-02-06 02:36:00,534 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44545.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:36:00,596 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44545.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:36:14,411 INFO [train.py:901] (1/4) Epoch 6, batch 4150, loss[loss=0.2636, simple_loss=0.3322, pruned_loss=0.09744, over 8106.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.3402, pruned_loss=0.1052, over 1615747.68 frames. ], batch size: 23, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:36:15,789 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44568.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:36:31,824 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6564, 2.3930, 4.5502, 1.3284, 2.9501, 2.1722, 1.6935, 2.5901], + device='cuda:1'), covar=tensor([0.1466, 0.1855, 0.0498, 0.3259, 0.1473, 0.2349, 0.1524, 0.2222], + device='cuda:1'), in_proj_covar=tensor([0.0466, 0.0455, 0.0530, 0.0533, 0.0581, 0.0520, 0.0441, 0.0583], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 02:36:49,812 INFO [train.py:901] (1/4) Epoch 6, batch 4200, loss[loss=0.2613, simple_loss=0.3251, pruned_loss=0.09873, over 8246.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3394, pruned_loss=0.1044, over 1615862.38 frames. ], batch size: 22, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:36:58,987 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.001e+02 2.806e+02 3.559e+02 4.787e+02 1.284e+03, threshold=7.119e+02, percent-clipped=4.0 +2023-02-06 02:37:05,647 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 02:37:07,969 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44642.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:37:12,004 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6756, 2.2348, 4.3144, 1.2527, 3.0079, 2.1104, 1.6813, 2.5107], + device='cuda:1'), covar=tensor([0.1396, 0.1868, 0.0584, 0.3176, 0.1203, 0.2322, 0.1495, 0.2161], + device='cuda:1'), in_proj_covar=tensor([0.0462, 0.0452, 0.0530, 0.0531, 0.0578, 0.0518, 0.0438, 0.0582], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 02:37:23,786 INFO [train.py:901] (1/4) Epoch 6, batch 4250, loss[loss=0.321, simple_loss=0.3839, pruned_loss=0.1291, over 8666.00 frames. ], tot_loss[loss=0.2742, simple_loss=0.34, pruned_loss=0.1042, over 1620678.48 frames. ], batch size: 34, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:37:24,684 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:37:29,255 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 02:37:41,572 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44692.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:37:58,491 INFO [train.py:901] (1/4) Epoch 6, batch 4300, loss[loss=0.2263, simple_loss=0.297, pruned_loss=0.07782, over 8253.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3404, pruned_loss=0.1044, over 1625418.54 frames. ], batch size: 22, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:38:00,033 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44717.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:38:08,670 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.771e+02 3.321e+02 4.102e+02 9.930e+02, threshold=6.641e+02, percent-clipped=2.0 +2023-02-06 02:38:17,344 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2498, 1.1578, 4.4641, 1.6473, 3.7990, 3.5745, 3.9541, 3.8683], + device='cuda:1'), covar=tensor([0.0485, 0.4139, 0.0376, 0.2791, 0.1109, 0.0695, 0.0511, 0.0578], + device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0504, 0.0456, 0.0442, 0.0512, 0.0421, 0.0422, 0.0478], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 02:38:33,192 INFO [train.py:901] (1/4) Epoch 6, batch 4350, loss[loss=0.2384, simple_loss=0.3041, pruned_loss=0.08633, over 8028.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3401, pruned_loss=0.1041, over 1618534.55 frames. ], batch size: 22, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:39:00,031 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 02:39:06,572 INFO [train.py:901] (1/4) Epoch 6, batch 4400, loss[loss=0.2406, simple_loss=0.3062, pruned_loss=0.08752, over 7660.00 frames. ], tot_loss[loss=0.276, simple_loss=0.3414, pruned_loss=0.1053, over 1616636.74 frames. ], batch size: 19, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:39:17,273 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.185e+02 3.434e+02 4.206e+02 5.183e+02 1.151e+03, threshold=8.413e+02, percent-clipped=11.0 +2023-02-06 02:39:40,225 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 02:39:42,260 INFO [train.py:901] (1/4) Epoch 6, batch 4450, loss[loss=0.2953, simple_loss=0.3689, pruned_loss=0.1108, over 8106.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3429, pruned_loss=0.1068, over 1614709.98 frames. ], batch size: 23, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:39:58,544 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44889.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:40:13,884 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44912.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:40:16,361 INFO [train.py:901] (1/4) Epoch 6, batch 4500, loss[loss=0.3337, simple_loss=0.3848, pruned_loss=0.1413, over 8724.00 frames. ], tot_loss[loss=0.279, simple_loss=0.3434, pruned_loss=0.1073, over 1615664.28 frames. ], batch size: 40, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:40:26,436 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.070e+02 3.100e+02 3.740e+02 5.266e+02 1.703e+03, threshold=7.479e+02, percent-clipped=4.0 +2023-02-06 02:40:31,754 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 02:40:48,305 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8577, 4.1284, 2.3492, 2.8753, 3.0583, 1.9799, 2.4404, 3.1552], + device='cuda:1'), covar=tensor([0.1301, 0.0236, 0.0843, 0.0590, 0.0599, 0.1079, 0.0940, 0.0671], + device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0235, 0.0307, 0.0300, 0.0313, 0.0311, 0.0332, 0.0302], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 02:40:52,040 INFO [train.py:901] (1/4) Epoch 6, batch 4550, loss[loss=0.3376, simple_loss=0.3795, pruned_loss=0.1479, over 8251.00 frames. ], tot_loss[loss=0.279, simple_loss=0.3435, pruned_loss=0.1073, over 1616261.22 frames. ], batch size: 22, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:41:18,811 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45004.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:41:27,066 INFO [train.py:901] (1/4) Epoch 6, batch 4600, loss[loss=0.3126, simple_loss=0.3569, pruned_loss=0.1342, over 6816.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.3413, pruned_loss=0.1059, over 1612352.36 frames. ], batch size: 71, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:41:34,845 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45027.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:41:36,664 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.977e+02 3.732e+02 4.434e+02 1.135e+03, threshold=7.465e+02, percent-clipped=1.0 +2023-02-06 02:42:02,768 INFO [train.py:901] (1/4) Epoch 6, batch 4650, loss[loss=0.3159, simple_loss=0.3613, pruned_loss=0.1352, over 7979.00 frames. ], tot_loss[loss=0.2771, simple_loss=0.3417, pruned_loss=0.1062, over 1615619.31 frames. ], batch size: 21, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:42:08,375 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45074.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:42:20,838 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 +2023-02-06 02:42:25,261 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45099.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:42:36,979 INFO [train.py:901] (1/4) Epoch 6, batch 4700, loss[loss=0.2138, simple_loss=0.2918, pruned_loss=0.06791, over 8095.00 frames. ], tot_loss[loss=0.2773, simple_loss=0.3414, pruned_loss=0.1066, over 1610665.60 frames. ], batch size: 21, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:42:46,395 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.041e+02 3.187e+02 3.833e+02 4.569e+02 1.251e+03, threshold=7.667e+02, percent-clipped=2.0 +2023-02-06 02:43:11,112 INFO [train.py:901] (1/4) Epoch 6, batch 4750, loss[loss=0.2708, simple_loss=0.3337, pruned_loss=0.1039, over 7541.00 frames. ], tot_loss[loss=0.2767, simple_loss=0.3413, pruned_loss=0.1061, over 1613751.22 frames. ], batch size: 18, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:43:30,431 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 02:43:31,820 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 02:43:46,228 INFO [train.py:901] (1/4) Epoch 6, batch 4800, loss[loss=0.2106, simple_loss=0.274, pruned_loss=0.07357, over 7708.00 frames. ], tot_loss[loss=0.2774, simple_loss=0.3418, pruned_loss=0.1065, over 1612061.14 frames. ], batch size: 18, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:43:55,770 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 3.021e+02 3.501e+02 4.623e+02 8.497e+02, threshold=7.001e+02, percent-clipped=1.0 +2023-02-06 02:44:16,235 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45260.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:44:20,163 INFO [train.py:901] (1/4) Epoch 6, batch 4850, loss[loss=0.3135, simple_loss=0.3682, pruned_loss=0.1294, over 8595.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3403, pruned_loss=0.106, over 1605138.49 frames. ], batch size: 39, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:44:20,859 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 02:44:22,416 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3639, 2.0964, 3.5136, 2.9848, 3.0786, 1.7422, 1.7770, 2.1561], + device='cuda:1'), covar=tensor([0.2211, 0.2665, 0.0634, 0.1183, 0.1155, 0.1723, 0.1322, 0.2181], + device='cuda:1'), in_proj_covar=tensor([0.0789, 0.0728, 0.0624, 0.0719, 0.0825, 0.0667, 0.0631, 0.0665], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:44:32,443 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:44:35,168 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45285.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:44:50,593 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45308.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:44:56,004 INFO [train.py:901] (1/4) Epoch 6, batch 4900, loss[loss=0.2946, simple_loss=0.3557, pruned_loss=0.1167, over 8133.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3408, pruned_loss=0.1065, over 1605035.60 frames. ], batch size: 22, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:45:04,267 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.9700, 1.4347, 4.2533, 1.6374, 3.6194, 3.4616, 3.7657, 3.6022], + device='cuda:1'), covar=tensor([0.0565, 0.3596, 0.0411, 0.2629, 0.1064, 0.0687, 0.0527, 0.0622], + device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0514, 0.0456, 0.0444, 0.0509, 0.0416, 0.0426, 0.0476], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 02:45:05,436 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.764e+02 2.896e+02 3.521e+02 4.501e+02 9.960e+02, threshold=7.042e+02, percent-clipped=7.0 +2023-02-06 02:45:06,568 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-06 02:45:13,523 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0802, 1.0225, 3.2655, 0.9355, 2.7183, 2.6483, 2.8826, 2.7736], + device='cuda:1'), covar=tensor([0.0694, 0.3816, 0.0611, 0.3054, 0.1465, 0.0925, 0.0703, 0.0817], + device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0515, 0.0455, 0.0443, 0.0511, 0.0417, 0.0427, 0.0475], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 02:45:30,239 INFO [train.py:901] (1/4) Epoch 6, batch 4950, loss[loss=0.2976, simple_loss=0.3556, pruned_loss=0.1198, over 8450.00 frames. ], tot_loss[loss=0.2784, simple_loss=0.3421, pruned_loss=0.1074, over 1609331.95 frames. ], batch size: 29, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:45:30,380 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45366.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:45:35,783 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3933, 1.3588, 2.7433, 1.2177, 2.0051, 3.0110, 2.9722, 2.5960], + device='cuda:1'), covar=tensor([0.1128, 0.1312, 0.0477, 0.2026, 0.0741, 0.0313, 0.0520, 0.0640], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0263, 0.0224, 0.0265, 0.0231, 0.0211, 0.0245, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 02:45:43,178 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1427, 0.9683, 3.2996, 0.9162, 2.7882, 2.6798, 2.9292, 2.8117], + device='cuda:1'), covar=tensor([0.0657, 0.3819, 0.0640, 0.3004, 0.1480, 0.0938, 0.0701, 0.0829], + device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0511, 0.0453, 0.0443, 0.0513, 0.0418, 0.0428, 0.0470], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 02:45:48,540 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3473, 1.8132, 3.0156, 2.2674, 2.4985, 1.9726, 1.6310, 1.0092], + device='cuda:1'), covar=tensor([0.2168, 0.2618, 0.0564, 0.1490, 0.1223, 0.1352, 0.1211, 0.2638], + device='cuda:1'), in_proj_covar=tensor([0.0789, 0.0731, 0.0622, 0.0717, 0.0824, 0.0670, 0.0636, 0.0667], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:45:54,531 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7800, 1.9442, 1.6627, 2.3908, 0.9703, 1.3311, 1.6758, 1.9627], + device='cuda:1'), covar=tensor([0.0983, 0.1094, 0.1446, 0.0590, 0.1656, 0.2286, 0.1305, 0.0938], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0242, 0.0277, 0.0220, 0.0240, 0.0273, 0.0279, 0.0246], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 02:46:05,709 INFO [train.py:901] (1/4) Epoch 6, batch 5000, loss[loss=0.2857, simple_loss=0.3584, pruned_loss=0.1065, over 8460.00 frames. ], tot_loss[loss=0.2774, simple_loss=0.3414, pruned_loss=0.1067, over 1608198.24 frames. ], batch size: 27, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:46:07,201 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45418.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:46:15,105 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.151e+02 3.255e+02 4.005e+02 4.887e+02 1.315e+03, threshold=8.009e+02, percent-clipped=7.0 +2023-02-06 02:46:24,036 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45443.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:46:40,014 INFO [train.py:901] (1/4) Epoch 6, batch 5050, loss[loss=0.3084, simple_loss=0.3729, pruned_loss=0.122, over 8520.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3424, pruned_loss=0.107, over 1611915.51 frames. ], batch size: 26, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:46:58,883 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 02:47:14,060 INFO [train.py:901] (1/4) Epoch 6, batch 5100, loss[loss=0.2945, simple_loss=0.3454, pruned_loss=0.1218, over 8029.00 frames. ], tot_loss[loss=0.2786, simple_loss=0.3429, pruned_loss=0.1071, over 1613835.38 frames. ], batch size: 22, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:47:24,716 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 2.842e+02 3.419e+02 4.219e+02 7.828e+02, threshold=6.837e+02, percent-clipped=0.0 +2023-02-06 02:47:26,314 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4146, 1.5011, 1.5350, 1.1194, 0.8928, 1.6674, 0.1630, 1.1104], + device='cuda:1'), covar=tensor([0.2794, 0.2100, 0.1003, 0.2190, 0.5405, 0.0560, 0.3769, 0.1954], + device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0136, 0.0087, 0.0184, 0.0228, 0.0083, 0.0147, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:47:26,970 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:47:29,627 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7001, 1.6539, 3.1770, 1.3676, 2.2033, 3.5280, 3.4641, 2.9756], + device='cuda:1'), covar=tensor([0.1071, 0.1189, 0.0363, 0.1790, 0.0730, 0.0262, 0.0473, 0.0611], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0263, 0.0226, 0.0264, 0.0231, 0.0212, 0.0248, 0.0278], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 02:47:29,672 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7670, 1.9699, 2.0845, 1.5226, 1.0338, 2.1690, 0.2786, 1.3175], + device='cuda:1'), covar=tensor([0.3084, 0.1657, 0.1019, 0.2524, 0.6026, 0.0514, 0.4649, 0.2240], + device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0136, 0.0088, 0.0185, 0.0228, 0.0083, 0.0148, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:47:43,452 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45558.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:47:49,317 INFO [train.py:901] (1/4) Epoch 6, batch 5150, loss[loss=0.2951, simple_loss=0.3339, pruned_loss=0.1282, over 8095.00 frames. ], tot_loss[loss=0.2802, simple_loss=0.344, pruned_loss=0.1082, over 1615113.68 frames. ], batch size: 21, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:48:23,753 INFO [train.py:901] (1/4) Epoch 6, batch 5200, loss[loss=0.2342, simple_loss=0.2997, pruned_loss=0.0844, over 7261.00 frames. ], tot_loss[loss=0.2799, simple_loss=0.3437, pruned_loss=0.108, over 1614428.61 frames. ], batch size: 16, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:48:34,006 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.204e+02 4.015e+02 4.654e+02 8.708e+02, threshold=8.029e+02, percent-clipped=4.0 +2023-02-06 02:48:37,738 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9445, 3.4099, 2.7521, 4.2148, 1.9381, 2.1587, 2.5229, 3.3173], + device='cuda:1'), covar=tensor([0.0777, 0.0814, 0.1093, 0.0293, 0.1301, 0.1695, 0.1410, 0.0874], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0244, 0.0278, 0.0224, 0.0240, 0.0275, 0.0282, 0.0248], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 02:48:48,646 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 02:48:57,682 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 02:48:59,756 INFO [train.py:901] (1/4) Epoch 6, batch 5250, loss[loss=0.3007, simple_loss=0.3595, pruned_loss=0.121, over 8346.00 frames. ], tot_loss[loss=0.2794, simple_loss=0.3428, pruned_loss=0.108, over 1612708.29 frames. ], batch size: 26, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:49:02,753 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2179, 1.5172, 2.2793, 1.0967, 1.6992, 1.4945, 1.3592, 1.2569], + device='cuda:1'), covar=tensor([0.1546, 0.1836, 0.0710, 0.3212, 0.1279, 0.2492, 0.1550, 0.1715], + device='cuda:1'), in_proj_covar=tensor([0.0466, 0.0464, 0.0531, 0.0540, 0.0586, 0.0526, 0.0443, 0.0588], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 02:49:30,211 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45710.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:49:34,107 INFO [train.py:901] (1/4) Epoch 6, batch 5300, loss[loss=0.273, simple_loss=0.3373, pruned_loss=0.1043, over 8236.00 frames. ], tot_loss[loss=0.2787, simple_loss=0.3428, pruned_loss=0.1073, over 1616515.16 frames. ], batch size: 22, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:49:36,949 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45720.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:49:43,555 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.559e+02 2.935e+02 3.437e+02 4.667e+02 1.283e+03, threshold=6.874e+02, percent-clipped=3.0 +2023-02-06 02:50:09,985 INFO [train.py:901] (1/4) Epoch 6, batch 5350, loss[loss=0.2723, simple_loss=0.3545, pruned_loss=0.09505, over 8295.00 frames. ], tot_loss[loss=0.2789, simple_loss=0.3433, pruned_loss=0.1073, over 1620736.32 frames. ], batch size: 23, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:50:25,433 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:50:27,332 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45792.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:50:42,754 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45814.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:50:42,772 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45814.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:50:43,901 INFO [train.py:901] (1/4) Epoch 6, batch 5400, loss[loss=0.2416, simple_loss=0.3245, pruned_loss=0.07933, over 8299.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3425, pruned_loss=0.107, over 1616726.54 frames. ], batch size: 23, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:50:49,017 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 02:50:49,970 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45825.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:50:53,692 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.842e+02 3.609e+02 4.644e+02 1.367e+03, threshold=7.218e+02, percent-clipped=2.0 +2023-02-06 02:50:59,042 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45839.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:51:16,139 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5044, 1.9332, 2.1288, 1.0555, 2.1836, 1.3883, 0.5072, 1.7156], + device='cuda:1'), covar=tensor([0.0263, 0.0144, 0.0107, 0.0221, 0.0144, 0.0451, 0.0356, 0.0120], + device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0247, 0.0207, 0.0298, 0.0237, 0.0384, 0.0308, 0.0282], + device='cuda:1'), out_proj_covar=tensor([1.1068e-04, 8.0056e-05, 6.6763e-05, 9.7360e-05, 7.8224e-05, 1.3616e-04, + 1.0250e-04, 9.2662e-05], device='cuda:1') +2023-02-06 02:51:17,291 INFO [train.py:901] (1/4) Epoch 6, batch 5450, loss[loss=0.3522, simple_loss=0.3933, pruned_loss=0.1555, over 6696.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3403, pruned_loss=0.1057, over 1616106.98 frames. ], batch size: 71, lr: 1.25e-02, grad_scale: 8.0 +2023-02-06 02:51:25,559 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45877.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:51:47,555 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 02:51:52,369 INFO [train.py:901] (1/4) Epoch 6, batch 5500, loss[loss=0.3079, simple_loss=0.3592, pruned_loss=0.1283, over 8289.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3399, pruned_loss=0.1058, over 1608624.73 frames. ], batch size: 23, lr: 1.25e-02, grad_scale: 8.0 +2023-02-06 02:52:03,083 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.821e+02 3.418e+02 4.385e+02 9.516e+02, threshold=6.836e+02, percent-clipped=4.0 +2023-02-06 02:52:05,947 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6776, 1.5491, 3.1963, 1.3023, 2.1644, 3.3980, 3.5802, 2.8687], + device='cuda:1'), covar=tensor([0.1094, 0.1352, 0.0389, 0.2024, 0.0824, 0.0386, 0.0358, 0.0761], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0262, 0.0224, 0.0263, 0.0232, 0.0212, 0.0247, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 02:52:27,011 INFO [train.py:901] (1/4) Epoch 6, batch 5550, loss[loss=0.2586, simple_loss=0.339, pruned_loss=0.08914, over 8295.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3399, pruned_loss=0.1051, over 1609584.88 frames. ], batch size: 23, lr: 1.25e-02, grad_scale: 8.0 +2023-02-06 02:53:03,503 INFO [train.py:901] (1/4) Epoch 6, batch 5600, loss[loss=0.2799, simple_loss=0.3475, pruned_loss=0.1061, over 8479.00 frames. ], tot_loss[loss=0.2743, simple_loss=0.3392, pruned_loss=0.1047, over 1609280.09 frames. ], batch size: 49, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:53:13,368 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.150e+02 2.809e+02 3.495e+02 4.670e+02 1.291e+03, threshold=6.989e+02, percent-clipped=6.0 +2023-02-06 02:53:36,015 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46064.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:53:37,305 INFO [train.py:901] (1/4) Epoch 6, batch 5650, loss[loss=0.2694, simple_loss=0.3389, pruned_loss=0.09991, over 8562.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.34, pruned_loss=0.1053, over 1609082.08 frames. ], batch size: 39, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:53:47,399 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46081.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:53:51,176 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 02:54:04,759 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46106.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:54:12,531 INFO [train.py:901] (1/4) Epoch 6, batch 5700, loss[loss=0.2667, simple_loss=0.3409, pruned_loss=0.09629, over 8027.00 frames. ], tot_loss[loss=0.2744, simple_loss=0.3395, pruned_loss=0.1047, over 1606881.79 frames. ], batch size: 22, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:54:22,542 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.829e+02 3.489e+02 4.392e+02 1.030e+03, threshold=6.978e+02, percent-clipped=3.0 +2023-02-06 02:54:25,975 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46136.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:54:34,077 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.6956, 1.1718, 3.9452, 1.2562, 3.3903, 3.2494, 3.5245, 3.4421], + device='cuda:1'), covar=tensor([0.0555, 0.3593, 0.0438, 0.2937, 0.1196, 0.0776, 0.0549, 0.0636], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0506, 0.0457, 0.0443, 0.0501, 0.0416, 0.0427, 0.0470], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 02:54:38,098 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.6903, 1.2964, 4.1445, 1.7308, 2.9690, 3.2389, 3.5955, 3.7259], + device='cuda:1'), covar=tensor([0.1268, 0.5706, 0.0927, 0.3647, 0.2438, 0.1430, 0.1237, 0.1121], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0505, 0.0456, 0.0441, 0.0500, 0.0415, 0.0426, 0.0468], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-02-06 02:54:46,327 INFO [train.py:901] (1/4) Epoch 6, batch 5750, loss[loss=0.2644, simple_loss=0.3369, pruned_loss=0.09596, over 8327.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3403, pruned_loss=0.1053, over 1614684.12 frames. ], batch size: 26, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:54:46,529 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9880, 3.9500, 2.6894, 2.5873, 2.9130, 2.1038, 2.6900, 2.9775], + device='cuda:1'), covar=tensor([0.1252, 0.0179, 0.0694, 0.0696, 0.0600, 0.0993, 0.0837, 0.0853], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0232, 0.0312, 0.0305, 0.0314, 0.0314, 0.0339, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 02:54:47,224 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8129, 2.7703, 3.2992, 0.8321, 3.3805, 1.9197, 1.4041, 1.9120], + device='cuda:1'), covar=tensor([0.0462, 0.0122, 0.0085, 0.0346, 0.0123, 0.0421, 0.0518, 0.0201], + device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0245, 0.0206, 0.0298, 0.0237, 0.0384, 0.0305, 0.0284], + device='cuda:1'), out_proj_covar=tensor([1.1131e-04, 7.9433e-05, 6.5978e-05, 9.6942e-05, 7.7894e-05, 1.3547e-04, + 1.0132e-04, 9.3017e-05], device='cuda:1') +2023-02-06 02:54:53,637 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 02:54:55,243 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46179.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:55:09,101 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9918, 2.6223, 2.9239, 1.4002, 3.1452, 1.9028, 1.4907, 1.8391], + device='cuda:1'), covar=tensor([0.0434, 0.0141, 0.0170, 0.0321, 0.0197, 0.0419, 0.0421, 0.0214], + device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0245, 0.0206, 0.0297, 0.0238, 0.0382, 0.0304, 0.0282], + device='cuda:1'), out_proj_covar=tensor([1.1058e-04, 7.9316e-05, 6.6154e-05, 9.6550e-05, 7.8107e-05, 1.3455e-04, + 1.0076e-04, 9.2262e-05], device='cuda:1') +2023-02-06 02:55:21,179 INFO [train.py:901] (1/4) Epoch 6, batch 5800, loss[loss=0.309, simple_loss=0.3671, pruned_loss=0.1255, over 8242.00 frames. ], tot_loss[loss=0.2771, simple_loss=0.3415, pruned_loss=0.1063, over 1610352.23 frames. ], batch size: 22, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:55:24,802 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46221.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:55:32,627 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.949e+02 3.358e+02 4.338e+02 9.471e+02, threshold=6.717e+02, percent-clipped=1.0 +2023-02-06 02:55:45,820 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46251.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:55:55,787 INFO [train.py:901] (1/4) Epoch 6, batch 5850, loss[loss=0.2801, simple_loss=0.3411, pruned_loss=0.1095, over 8129.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3413, pruned_loss=0.1062, over 1613012.50 frames. ], batch size: 22, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:56:14,982 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46294.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:56:29,478 INFO [train.py:901] (1/4) Epoch 6, batch 5900, loss[loss=0.2863, simple_loss=0.3318, pruned_loss=0.1204, over 7697.00 frames. ], tot_loss[loss=0.2771, simple_loss=0.3414, pruned_loss=0.1064, over 1613825.71 frames. ], batch size: 18, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:56:39,482 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 3.022e+02 3.849e+02 5.141e+02 8.536e+02, threshold=7.697e+02, percent-clipped=7.0 +2023-02-06 02:56:43,692 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:57:04,148 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6443, 1.3165, 1.2953, 1.1342, 0.9426, 1.1597, 1.3115, 1.3732], + device='cuda:1'), covar=tensor([0.0514, 0.1089, 0.1456, 0.1216, 0.0548, 0.1390, 0.0675, 0.0483], + device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0172, 0.0212, 0.0175, 0.0122, 0.0182, 0.0136, 0.0148], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 02:57:04,663 INFO [train.py:901] (1/4) Epoch 6, batch 5950, loss[loss=0.211, simple_loss=0.2803, pruned_loss=0.07087, over 5542.00 frames. ], tot_loss[loss=0.2757, simple_loss=0.3408, pruned_loss=0.1054, over 1613638.93 frames. ], batch size: 12, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:57:38,437 INFO [train.py:901] (1/4) Epoch 6, batch 6000, loss[loss=0.2911, simple_loss=0.3608, pruned_loss=0.1107, over 7928.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.3402, pruned_loss=0.1046, over 1617598.54 frames. ], batch size: 20, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:57:38,437 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 02:57:50,762 INFO [train.py:935] (1/4) Epoch 6, validation: loss=0.2127, simple_loss=0.3094, pruned_loss=0.05799, over 944034.00 frames. +2023-02-06 02:57:50,763 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6607MB +2023-02-06 02:58:01,252 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.784e+02 2.683e+02 3.226e+02 4.100e+02 1.140e+03, threshold=6.453e+02, percent-clipped=1.0 +2023-02-06 02:58:04,363 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:58:21,780 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46460.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:58:25,656 INFO [train.py:901] (1/4) Epoch 6, batch 6050, loss[loss=0.2495, simple_loss=0.3281, pruned_loss=0.08549, over 8044.00 frames. ], tot_loss[loss=0.2732, simple_loss=0.3392, pruned_loss=0.1036, over 1619851.07 frames. ], batch size: 22, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:58:37,126 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 02:58:39,501 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0756, 3.0728, 2.8032, 1.4256, 2.6985, 2.8009, 2.8116, 2.6708], + device='cuda:1'), covar=tensor([0.1509, 0.0983, 0.1358, 0.5370, 0.1142, 0.1340, 0.1811, 0.1269], + device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0315, 0.0341, 0.0434, 0.0332, 0.0305, 0.0329, 0.0275], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 02:58:56,114 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:59:02,119 INFO [train.py:901] (1/4) Epoch 6, batch 6100, loss[loss=0.2443, simple_loss=0.3027, pruned_loss=0.09296, over 7789.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.3389, pruned_loss=0.1039, over 1615579.73 frames. ], batch size: 19, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:59:12,635 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 3.046e+02 3.657e+02 4.398e+02 9.620e+02, threshold=7.315e+02, percent-clipped=4.0 +2023-02-06 02:59:13,565 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:59:24,581 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 02:59:37,765 INFO [train.py:901] (1/4) Epoch 6, batch 6150, loss[loss=0.2779, simple_loss=0.3448, pruned_loss=0.1055, over 8572.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.3377, pruned_loss=0.103, over 1615796.32 frames. ], batch size: 31, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:59:56,277 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:09,154 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8107, 3.2137, 3.4564, 2.5602, 1.4642, 3.4689, 0.6602, 2.1785], + device='cuda:1'), covar=tensor([0.3457, 0.1585, 0.0696, 0.2830, 0.6547, 0.0617, 0.6133, 0.1869], + device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0133, 0.0082, 0.0182, 0.0219, 0.0080, 0.0143, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:00:11,879 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46613.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:13,823 INFO [train.py:901] (1/4) Epoch 6, batch 6200, loss[loss=0.3978, simple_loss=0.4223, pruned_loss=0.1867, over 7154.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3381, pruned_loss=0.1039, over 1611175.38 frames. ], batch size: 72, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 03:00:14,743 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46617.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:16,774 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:24,153 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.053e+02 3.051e+02 3.861e+02 4.926e+02 1.016e+03, threshold=7.722e+02, percent-clipped=3.0 +2023-02-06 03:00:28,971 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46638.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:48,686 INFO [train.py:901] (1/4) Epoch 6, batch 6250, loss[loss=0.2226, simple_loss=0.301, pruned_loss=0.07211, over 7942.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3365, pruned_loss=0.1029, over 1604893.11 frames. ], batch size: 20, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 03:01:22,862 INFO [train.py:901] (1/4) Epoch 6, batch 6300, loss[loss=0.2332, simple_loss=0.2996, pruned_loss=0.0834, over 8078.00 frames. ], tot_loss[loss=0.2708, simple_loss=0.3364, pruned_loss=0.1026, over 1604518.41 frames. ], batch size: 21, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 03:01:34,427 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.898e+02 2.734e+02 3.399e+02 4.377e+02 1.449e+03, threshold=6.797e+02, percent-clipped=4.0 +2023-02-06 03:01:49,385 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46753.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:01:57,779 INFO [train.py:901] (1/4) Epoch 6, batch 6350, loss[loss=0.3493, simple_loss=0.3984, pruned_loss=0.1501, over 8342.00 frames. ], tot_loss[loss=0.2719, simple_loss=0.3376, pruned_loss=0.1031, over 1607376.53 frames. ], batch size: 26, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:01:58,516 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.3322, 1.6473, 1.5772, 1.3180, 1.4180, 1.7383, 2.3296, 1.7531], + device='cuda:1'), covar=tensor([0.0416, 0.1316, 0.1743, 0.1368, 0.0609, 0.1554, 0.0633, 0.0624], + device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0169, 0.0210, 0.0172, 0.0120, 0.0178, 0.0133, 0.0147], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 03:02:32,255 INFO [train.py:901] (1/4) Epoch 6, batch 6400, loss[loss=0.2844, simple_loss=0.3522, pruned_loss=0.1083, over 8453.00 frames. ], tot_loss[loss=0.2715, simple_loss=0.3371, pruned_loss=0.1029, over 1611053.02 frames. ], batch size: 27, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:02:41,189 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46828.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:02:43,104 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.744e+02 3.578e+02 4.396e+02 9.504e+02, threshold=7.157e+02, percent-clipped=5.0 +2023-02-06 03:03:07,338 INFO [train.py:901] (1/4) Epoch 6, batch 6450, loss[loss=0.2593, simple_loss=0.3385, pruned_loss=0.09005, over 8453.00 frames. ], tot_loss[loss=0.2714, simple_loss=0.3371, pruned_loss=0.1028, over 1609943.81 frames. ], batch size: 27, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:03:41,565 INFO [train.py:901] (1/4) Epoch 6, batch 6500, loss[loss=0.2171, simple_loss=0.2822, pruned_loss=0.07595, over 7797.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3362, pruned_loss=0.103, over 1609733.68 frames. ], batch size: 19, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:03:51,600 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.994e+02 3.001e+02 3.759e+02 4.377e+02 1.086e+03, threshold=7.517e+02, percent-clipped=1.0 +2023-02-06 03:04:09,449 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46957.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:04:14,545 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46964.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:04:15,766 INFO [train.py:901] (1/4) Epoch 6, batch 6550, loss[loss=0.2595, simple_loss=0.3221, pruned_loss=0.09842, over 7434.00 frames. ], tot_loss[loss=0.2719, simple_loss=0.3363, pruned_loss=0.1037, over 1610830.81 frames. ], batch size: 17, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:04:37,515 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 03:04:45,799 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47009.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:04:50,982 INFO [train.py:901] (1/4) Epoch 6, batch 6600, loss[loss=0.2973, simple_loss=0.3648, pruned_loss=0.1149, over 8706.00 frames. ], tot_loss[loss=0.2708, simple_loss=0.3354, pruned_loss=0.1031, over 1608987.47 frames. ], batch size: 34, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:04:56,436 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 03:05:01,160 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.072e+02 2.927e+02 3.687e+02 4.772e+02 1.123e+03, threshold=7.374e+02, percent-clipped=4.0 +2023-02-06 03:05:03,300 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47034.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:05:25,535 INFO [train.py:901] (1/4) Epoch 6, batch 6650, loss[loss=0.2694, simple_loss=0.3234, pruned_loss=0.1077, over 7818.00 frames. ], tot_loss[loss=0.2719, simple_loss=0.3365, pruned_loss=0.1037, over 1605830.82 frames. ], batch size: 20, lr: 1.23e-02, grad_scale: 4.0 +2023-02-06 03:05:30,597 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47072.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:05:35,422 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47079.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:06:00,869 INFO [train.py:901] (1/4) Epoch 6, batch 6700, loss[loss=0.235, simple_loss=0.3116, pruned_loss=0.07919, over 8475.00 frames. ], tot_loss[loss=0.2714, simple_loss=0.3366, pruned_loss=0.1031, over 1611651.32 frames. ], batch size: 25, lr: 1.23e-02, grad_scale: 4.0 +2023-02-06 03:06:12,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.777e+02 3.640e+02 4.922e+02 1.093e+03, threshold=7.281e+02, percent-clipped=6.0 +2023-02-06 03:06:28,292 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4876, 1.8687, 1.5272, 2.4048, 1.2023, 1.2678, 1.7916, 1.9352], + device='cuda:1'), covar=tensor([0.1138, 0.1025, 0.1573, 0.0482, 0.1345, 0.1977, 0.1103, 0.1047], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0246, 0.0279, 0.0224, 0.0244, 0.0272, 0.0276, 0.0247], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 03:06:28,355 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1843, 1.7929, 2.6403, 2.0559, 2.2637, 1.9134, 1.4716, 0.9229], + device='cuda:1'), covar=tensor([0.2251, 0.2347, 0.0626, 0.1342, 0.1192, 0.1278, 0.1247, 0.2537], + device='cuda:1'), in_proj_covar=tensor([0.0799, 0.0735, 0.0639, 0.0724, 0.0825, 0.0677, 0.0634, 0.0677], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:06:34,847 INFO [train.py:901] (1/4) Epoch 6, batch 6750, loss[loss=0.2862, simple_loss=0.3605, pruned_loss=0.106, over 8104.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3365, pruned_loss=0.1027, over 1610999.99 frames. ], batch size: 23, lr: 1.23e-02, grad_scale: 4.0 +2023-02-06 03:06:38,956 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47172.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:07:01,070 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-02-06 03:07:10,710 INFO [train.py:901] (1/4) Epoch 6, batch 6800, loss[loss=0.2927, simple_loss=0.3606, pruned_loss=0.1124, over 8498.00 frames. ], tot_loss[loss=0.2707, simple_loss=0.336, pruned_loss=0.1026, over 1608081.24 frames. ], batch size: 26, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:07:12,696 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 03:07:21,207 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.942e+02 3.591e+02 4.804e+02 1.528e+03, threshold=7.182e+02, percent-clipped=7.0 +2023-02-06 03:07:37,978 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1364, 1.7051, 1.3556, 1.7119, 1.3666, 1.1187, 1.2630, 1.3796], + device='cuda:1'), covar=tensor([0.0924, 0.0348, 0.0927, 0.0455, 0.0613, 0.1209, 0.0840, 0.0711], + device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0230, 0.0307, 0.0299, 0.0308, 0.0311, 0.0338, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 03:07:40,024 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0148, 1.1418, 4.2734, 1.5186, 3.5959, 3.5373, 3.7716, 3.5845], + device='cuda:1'), covar=tensor([0.0580, 0.4306, 0.0412, 0.2955, 0.1178, 0.0806, 0.0532, 0.0694], + device='cuda:1'), in_proj_covar=tensor([0.0371, 0.0511, 0.0450, 0.0448, 0.0507, 0.0428, 0.0430, 0.0482], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 03:07:40,034 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47259.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:07:44,572 INFO [train.py:901] (1/4) Epoch 6, batch 6850, loss[loss=0.3871, simple_loss=0.4073, pruned_loss=0.1835, over 8282.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.338, pruned_loss=0.1037, over 1610625.79 frames. ], batch size: 23, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:07:58,966 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47287.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:08:00,906 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 03:08:19,245 INFO [train.py:901] (1/4) Epoch 6, batch 6900, loss[loss=0.2515, simple_loss=0.3179, pruned_loss=0.09255, over 8291.00 frames. ], tot_loss[loss=0.2725, simple_loss=0.3379, pruned_loss=0.1035, over 1611167.05 frames. ], batch size: 23, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:08:28,205 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:08:30,615 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 2.873e+02 3.537e+02 4.379e+02 9.664e+02, threshold=7.075e+02, percent-clipped=2.0 +2023-02-06 03:08:32,865 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47335.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:08:44,870 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47353.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:08:49,771 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47360.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:08:54,249 INFO [train.py:901] (1/4) Epoch 6, batch 6950, loss[loss=0.2036, simple_loss=0.2663, pruned_loss=0.07047, over 7426.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.338, pruned_loss=0.1037, over 1612816.38 frames. ], batch size: 17, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:09:09,777 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 03:09:15,147 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47397.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:09:28,488 INFO [train.py:901] (1/4) Epoch 6, batch 7000, loss[loss=0.2886, simple_loss=0.3664, pruned_loss=0.1055, over 8349.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3378, pruned_loss=0.1028, over 1618057.64 frames. ], batch size: 24, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:09:39,928 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.784e+02 3.553e+02 4.437e+02 1.281e+03, threshold=7.106e+02, percent-clipped=4.0 +2023-02-06 03:10:01,757 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4388, 2.0042, 3.3363, 2.5998, 2.6557, 1.9756, 1.4312, 1.2264], + device='cuda:1'), covar=tensor([0.2163, 0.2645, 0.0525, 0.1310, 0.1138, 0.1285, 0.1240, 0.2633], + device='cuda:1'), in_proj_covar=tensor([0.0814, 0.0741, 0.0647, 0.0737, 0.0840, 0.0689, 0.0646, 0.0686], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:10:03,576 INFO [train.py:901] (1/4) Epoch 6, batch 7050, loss[loss=0.2286, simple_loss=0.3114, pruned_loss=0.07289, over 8358.00 frames. ], tot_loss[loss=0.2725, simple_loss=0.3388, pruned_loss=0.1031, over 1619100.87 frames. ], batch size: 24, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:10:24,717 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6247, 1.8828, 2.0860, 1.5756, 0.9291, 2.2063, 0.3325, 1.3538], + device='cuda:1'), covar=tensor([0.3863, 0.1850, 0.0879, 0.2591, 0.5952, 0.0502, 0.4738, 0.2223], + device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0135, 0.0082, 0.0181, 0.0225, 0.0081, 0.0142, 0.0136], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:10:37,638 INFO [train.py:901] (1/4) Epoch 6, batch 7100, loss[loss=0.2415, simple_loss=0.3055, pruned_loss=0.08872, over 7660.00 frames. ], tot_loss[loss=0.2722, simple_loss=0.3381, pruned_loss=0.1031, over 1616096.87 frames. ], batch size: 19, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:10:48,827 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.154e+02 3.207e+02 3.842e+02 5.073e+02 1.424e+03, threshold=7.684e+02, percent-clipped=2.0 +2023-02-06 03:10:53,127 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6035, 1.9649, 3.4176, 1.1583, 2.3237, 1.8983, 1.6069, 1.9966], + device='cuda:1'), covar=tensor([0.1340, 0.1630, 0.0605, 0.3160, 0.1333, 0.2168, 0.1362, 0.2155], + device='cuda:1'), in_proj_covar=tensor([0.0467, 0.0466, 0.0535, 0.0548, 0.0590, 0.0527, 0.0446, 0.0595], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 03:10:56,274 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:11:12,591 INFO [train.py:901] (1/4) Epoch 6, batch 7150, loss[loss=0.3066, simple_loss=0.3657, pruned_loss=0.1237, over 8504.00 frames. ], tot_loss[loss=0.2736, simple_loss=0.3388, pruned_loss=0.1042, over 1607905.34 frames. ], batch size: 34, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:11:14,075 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47568.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:11:29,991 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 03:11:33,775 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9861, 2.4404, 2.6620, 1.0510, 2.7101, 1.4876, 1.4754, 1.6618], + device='cuda:1'), covar=tensor([0.0383, 0.0151, 0.0136, 0.0349, 0.0247, 0.0476, 0.0411, 0.0255], + device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0251, 0.0208, 0.0302, 0.0245, 0.0385, 0.0309, 0.0287], + device='cuda:1'), out_proj_covar=tensor([1.1073e-04, 8.0746e-05, 6.5994e-05, 9.7173e-05, 8.0053e-05, 1.3478e-04, + 1.0181e-04, 9.3043e-05], device='cuda:1') +2023-02-06 03:11:37,955 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:11:46,775 INFO [train.py:901] (1/4) Epoch 6, batch 7200, loss[loss=0.213, simple_loss=0.2893, pruned_loss=0.06841, over 7796.00 frames. ], tot_loss[loss=0.2744, simple_loss=0.3395, pruned_loss=0.1046, over 1609882.12 frames. ], batch size: 19, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:11:57,794 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.983e+02 3.737e+02 4.630e+02 8.445e+02, threshold=7.473e+02, percent-clipped=4.0 +2023-02-06 03:12:22,027 INFO [train.py:901] (1/4) Epoch 6, batch 7250, loss[loss=0.2513, simple_loss=0.3296, pruned_loss=0.08655, over 8355.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.3393, pruned_loss=0.1036, over 1614805.36 frames. ], batch size: 24, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:12:56,482 INFO [train.py:901] (1/4) Epoch 6, batch 7300, loss[loss=0.2971, simple_loss=0.3581, pruned_loss=0.118, over 7974.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3398, pruned_loss=0.1042, over 1617926.85 frames. ], batch size: 21, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:12:57,904 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:13:07,203 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 3.071e+02 3.696e+02 4.839e+02 1.031e+03, threshold=7.393e+02, percent-clipped=2.0 +2023-02-06 03:13:13,260 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47741.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:13:23,405 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47756.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:13:30,054 INFO [train.py:901] (1/4) Epoch 6, batch 7350, loss[loss=0.2725, simple_loss=0.3176, pruned_loss=0.1137, over 5937.00 frames. ], tot_loss[loss=0.2732, simple_loss=0.339, pruned_loss=0.1037, over 1615578.84 frames. ], batch size: 13, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:13:45,224 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 03:13:48,789 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 03:13:59,324 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47806.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:14:05,872 INFO [train.py:901] (1/4) Epoch 6, batch 7400, loss[loss=0.2554, simple_loss=0.3209, pruned_loss=0.09493, over 7806.00 frames. ], tot_loss[loss=0.2743, simple_loss=0.3398, pruned_loss=0.1044, over 1614695.05 frames. ], batch size: 20, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:14:08,003 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 03:14:13,529 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47827.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:14:17,407 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 3.124e+02 3.904e+02 4.877e+02 9.892e+02, threshold=7.808e+02, percent-clipped=5.0 +2023-02-06 03:14:33,728 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:14:40,242 INFO [train.py:901] (1/4) Epoch 6, batch 7450, loss[loss=0.2441, simple_loss=0.3093, pruned_loss=0.08942, over 7815.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.3405, pruned_loss=0.105, over 1610445.95 frames. ], batch size: 20, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:14:46,244 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 03:15:15,323 INFO [train.py:901] (1/4) Epoch 6, batch 7500, loss[loss=0.2405, simple_loss=0.3189, pruned_loss=0.08105, over 8527.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.3402, pruned_loss=0.1046, over 1614283.56 frames. ], batch size: 28, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:15:25,968 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.098e+02 3.102e+02 3.706e+02 4.699e+02 1.511e+03, threshold=7.412e+02, percent-clipped=9.0 +2023-02-06 03:15:49,279 INFO [train.py:901] (1/4) Epoch 6, batch 7550, loss[loss=0.325, simple_loss=0.3857, pruned_loss=0.1321, over 8517.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3407, pruned_loss=0.1043, over 1618747.76 frames. ], batch size: 26, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:15:54,925 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47974.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:16:11,852 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:16:24,741 INFO [train.py:901] (1/4) Epoch 6, batch 7600, loss[loss=0.2094, simple_loss=0.2762, pruned_loss=0.0713, over 7928.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.3405, pruned_loss=0.1045, over 1619893.62 frames. ], batch size: 20, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:16:37,193 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.897e+02 3.536e+02 4.611e+02 2.294e+03, threshold=7.072e+02, percent-clipped=5.0 +2023-02-06 03:17:01,520 INFO [train.py:901] (1/4) Epoch 6, batch 7650, loss[loss=0.2886, simple_loss=0.3521, pruned_loss=0.1126, over 7176.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.341, pruned_loss=0.1043, over 1624470.75 frames. ], batch size: 72, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:17:03,782 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1282, 2.4285, 2.0016, 2.9416, 1.5232, 1.5501, 2.0946, 2.6943], + device='cuda:1'), covar=tensor([0.1051, 0.1178, 0.1283, 0.0567, 0.1440, 0.2065, 0.1245, 0.0947], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0248, 0.0277, 0.0229, 0.0246, 0.0272, 0.0278, 0.0248], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 03:17:17,563 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48090.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:17:24,158 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48100.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:17:32,462 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48112.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:17:34,944 INFO [train.py:901] (1/4) Epoch 6, batch 7700, loss[loss=0.2488, simple_loss=0.3339, pruned_loss=0.08179, over 8183.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3385, pruned_loss=0.103, over 1621096.81 frames. ], batch size: 23, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:17:45,628 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6714, 3.1394, 3.2575, 2.2552, 1.3705, 3.2766, 0.4806, 2.1256], + device='cuda:1'), covar=tensor([0.2493, 0.1110, 0.0754, 0.2976, 0.6538, 0.0663, 0.5229, 0.2414], + device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0132, 0.0081, 0.0183, 0.0231, 0.0082, 0.0143, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:17:46,047 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.945e+02 2.821e+02 3.617e+02 4.667e+02 9.808e+02, threshold=7.234e+02, percent-clipped=3.0 +2023-02-06 03:17:50,863 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48137.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:17:57,317 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 03:17:59,320 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48150.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:18:10,116 INFO [train.py:901] (1/4) Epoch 6, batch 7750, loss[loss=0.2376, simple_loss=0.2951, pruned_loss=0.09007, over 7533.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3383, pruned_loss=0.103, over 1622898.80 frames. ], batch size: 18, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:18:13,408 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48171.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:18:18,148 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4480, 1.8954, 2.1016, 0.9393, 2.2129, 1.3734, 0.5106, 1.7968], + device='cuda:1'), covar=tensor([0.0303, 0.0158, 0.0117, 0.0259, 0.0162, 0.0428, 0.0396, 0.0140], + device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0251, 0.0206, 0.0304, 0.0244, 0.0388, 0.0319, 0.0290], + device='cuda:1'), out_proj_covar=tensor([1.1360e-04, 8.0534e-05, 6.4772e-05, 9.7039e-05, 7.9117e-05, 1.3542e-04, + 1.0453e-04, 9.3706e-05], device='cuda:1') +2023-02-06 03:18:28,105 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9788, 1.4155, 4.3670, 1.7483, 2.3567, 5.0622, 4.8316, 4.3102], + device='cuda:1'), covar=tensor([0.1203, 0.1608, 0.0298, 0.2132, 0.0867, 0.0211, 0.0352, 0.0613], + device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0272, 0.0227, 0.0273, 0.0235, 0.0212, 0.0260, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 03:18:39,663 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-06 03:18:39,964 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5888, 2.0262, 3.4701, 1.2823, 2.5385, 1.8722, 1.6819, 2.0844], + device='cuda:1'), covar=tensor([0.1438, 0.1688, 0.0574, 0.2976, 0.1250, 0.2252, 0.1464, 0.2027], + device='cuda:1'), in_proj_covar=tensor([0.0468, 0.0466, 0.0530, 0.0549, 0.0595, 0.0532, 0.0449, 0.0592], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 03:18:43,312 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48215.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:18:43,803 INFO [train.py:901] (1/4) Epoch 6, batch 7800, loss[loss=0.2435, simple_loss=0.3275, pruned_loss=0.07978, over 8363.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.338, pruned_loss=0.1027, over 1619571.39 frames. ], batch size: 24, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:18:53,439 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48230.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:18:54,591 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.145e+02 3.053e+02 3.731e+02 4.789e+02 1.133e+03, threshold=7.462e+02, percent-clipped=3.0 +2023-02-06 03:19:16,578 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48265.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:19:17,073 INFO [train.py:901] (1/4) Epoch 6, batch 7850, loss[loss=0.2636, simple_loss=0.3274, pruned_loss=0.09985, over 7812.00 frames. ], tot_loss[loss=0.2703, simple_loss=0.3365, pruned_loss=0.102, over 1615540.32 frames. ], batch size: 20, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:19:30,601 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48286.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:19:51,010 INFO [train.py:901] (1/4) Epoch 6, batch 7900, loss[loss=0.2717, simple_loss=0.3434, pruned_loss=0.1001, over 8731.00 frames. ], tot_loss[loss=0.2722, simple_loss=0.3376, pruned_loss=0.1034, over 1614718.25 frames. ], batch size: 40, lr: 1.21e-02, grad_scale: 8.0 +2023-02-06 03:20:01,872 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.890e+02 2.927e+02 3.494e+02 4.326e+02 7.205e+02, threshold=6.988e+02, percent-clipped=0.0 +2023-02-06 03:20:09,328 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:20:25,102 INFO [train.py:901] (1/4) Epoch 6, batch 7950, loss[loss=0.2966, simple_loss=0.3576, pruned_loss=0.1178, over 8426.00 frames. ], tot_loss[loss=0.2707, simple_loss=0.3362, pruned_loss=0.1026, over 1612753.52 frames. ], batch size: 49, lr: 1.21e-02, grad_scale: 8.0 +2023-02-06 03:20:58,705 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48415.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:20:59,254 INFO [train.py:901] (1/4) Epoch 6, batch 8000, loss[loss=0.1898, simple_loss=0.2701, pruned_loss=0.05476, over 7700.00 frames. ], tot_loss[loss=0.2715, simple_loss=0.3369, pruned_loss=0.103, over 1612118.50 frames. ], batch size: 18, lr: 1.21e-02, grad_scale: 8.0 +2023-02-06 03:21:10,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.174e+02 2.873e+02 3.488e+02 4.217e+02 8.104e+02, threshold=6.977e+02, percent-clipped=2.0 +2023-02-06 03:21:11,768 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:21:16,771 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48441.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:21:33,924 INFO [train.py:901] (1/4) Epoch 6, batch 8050, loss[loss=0.1878, simple_loss=0.2561, pruned_loss=0.05975, over 6799.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3351, pruned_loss=0.103, over 1585351.63 frames. ], batch size: 15, lr: 1.21e-02, grad_scale: 8.0 +2023-02-06 03:21:37,661 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48471.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:21:54,561 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48496.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:22:07,122 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 03:22:10,943 INFO [train.py:901] (1/4) Epoch 7, batch 0, loss[loss=0.2877, simple_loss=0.3442, pruned_loss=0.1156, over 8132.00 frames. ], tot_loss[loss=0.2877, simple_loss=0.3442, pruned_loss=0.1156, over 8132.00 frames. ], batch size: 22, lr: 1.14e-02, grad_scale: 8.0 +2023-02-06 03:22:10,943 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 03:22:22,764 INFO [train.py:935] (1/4) Epoch 7, validation: loss=0.2113, simple_loss=0.3091, pruned_loss=0.05678, over 944034.00 frames. +2023-02-06 03:22:22,765 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6607MB +2023-02-06 03:22:28,411 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9532, 2.3022, 1.6967, 2.8375, 1.3141, 1.4630, 1.8004, 2.2859], + device='cuda:1'), covar=tensor([0.0901, 0.1105, 0.1438, 0.0440, 0.1419, 0.1789, 0.1298, 0.0952], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0239, 0.0276, 0.0221, 0.0240, 0.0267, 0.0273, 0.0244], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 03:22:37,630 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48521.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:22:38,086 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 03:22:39,037 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3442, 1.6656, 1.6303, 0.8153, 1.7435, 1.1872, 0.2628, 1.5441], + device='cuda:1'), covar=tensor([0.0216, 0.0149, 0.0114, 0.0213, 0.0146, 0.0414, 0.0381, 0.0112], + device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0251, 0.0203, 0.0302, 0.0238, 0.0385, 0.0310, 0.0287], + device='cuda:1'), out_proj_covar=tensor([1.1150e-04, 8.0551e-05, 6.3816e-05, 9.6298e-05, 7.6935e-05, 1.3405e-04, + 1.0118e-04, 9.2305e-05], device='cuda:1') +2023-02-06 03:22:41,676 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3224, 1.4703, 1.3324, 1.9170, 0.7715, 1.1810, 1.2681, 1.4537], + device='cuda:1'), covar=tensor([0.1247, 0.1140, 0.1648, 0.0676, 0.1552, 0.2006, 0.1218, 0.1082], + device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0242, 0.0278, 0.0224, 0.0244, 0.0271, 0.0277, 0.0247], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 03:22:45,430 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.727e+02 3.570e+02 4.321e+02 1.428e+03, threshold=7.140e+02, percent-clipped=5.0 +2023-02-06 03:22:53,288 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48542.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:22:55,822 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48546.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:22:56,685 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 03:22:57,635 INFO [train.py:901] (1/4) Epoch 7, batch 50, loss[loss=0.3065, simple_loss=0.359, pruned_loss=0.1271, over 8234.00 frames. ], tot_loss[loss=0.2713, simple_loss=0.3385, pruned_loss=0.1021, over 366194.36 frames. ], batch size: 22, lr: 1.14e-02, grad_scale: 8.0 +2023-02-06 03:22:57,790 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48549.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:23:09,796 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48567.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:23:12,934 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 03:23:14,282 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48574.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:23:31,385 INFO [train.py:901] (1/4) Epoch 7, batch 100, loss[loss=0.2815, simple_loss=0.3451, pruned_loss=0.1089, over 8675.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3383, pruned_loss=0.1019, over 646438.79 frames. ], batch size: 31, lr: 1.14e-02, grad_scale: 8.0 +2023-02-06 03:23:34,998 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 03:23:44,211 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5580, 2.0063, 3.4963, 1.2820, 2.5842, 2.1093, 1.7009, 2.0216], + device='cuda:1'), covar=tensor([0.1478, 0.1724, 0.0581, 0.3076, 0.1234, 0.2163, 0.1423, 0.2044], + device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0473, 0.0526, 0.0547, 0.0594, 0.0534, 0.0451, 0.0592], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 03:23:54,579 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.936e+02 3.434e+02 4.642e+02 8.961e+02, threshold=6.868e+02, percent-clipped=3.0 +2023-02-06 03:24:06,731 INFO [train.py:901] (1/4) Epoch 7, batch 150, loss[loss=0.2533, simple_loss=0.3169, pruned_loss=0.09485, over 7775.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3368, pruned_loss=0.1007, over 862404.77 frames. ], batch size: 19, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:24:28,700 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5855, 2.2160, 4.5705, 1.2990, 2.7021, 2.1801, 1.6328, 2.7617], + device='cuda:1'), covar=tensor([0.1675, 0.2100, 0.0643, 0.3527, 0.1806, 0.2570, 0.1682, 0.2466], + device='cuda:1'), in_proj_covar=tensor([0.0470, 0.0470, 0.0523, 0.0541, 0.0591, 0.0527, 0.0448, 0.0587], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 03:24:31,972 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48686.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:24:34,077 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48689.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:24:40,629 INFO [train.py:901] (1/4) Epoch 7, batch 200, loss[loss=0.2468, simple_loss=0.3267, pruned_loss=0.0835, over 8300.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3346, pruned_loss=0.0999, over 1024016.76 frames. ], batch size: 23, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:25:03,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 2.629e+02 3.306e+02 4.274e+02 1.004e+03, threshold=6.612e+02, percent-clipped=3.0 +2023-02-06 03:25:15,507 INFO [train.py:901] (1/4) Epoch 7, batch 250, loss[loss=0.2565, simple_loss=0.3268, pruned_loss=0.09313, over 8187.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.3332, pruned_loss=0.09864, over 1158267.80 frames. ], batch size: 23, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:25:22,707 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48759.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:25:26,762 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 03:25:35,605 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 03:25:41,109 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:25:44,863 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-06 03:25:50,548 INFO [train.py:901] (1/4) Epoch 7, batch 300, loss[loss=0.2947, simple_loss=0.3637, pruned_loss=0.1128, over 8187.00 frames. ], tot_loss[loss=0.2663, simple_loss=0.3335, pruned_loss=0.09957, over 1255185.02 frames. ], batch size: 23, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:25:52,221 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48801.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:25:54,987 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48805.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:26:12,383 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48830.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:26:13,531 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 2.973e+02 3.476e+02 4.340e+02 1.124e+03, threshold=6.953e+02, percent-clipped=5.0 +2023-02-06 03:26:18,393 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48839.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:26:25,097 INFO [train.py:901] (1/4) Epoch 7, batch 350, loss[loss=0.3251, simple_loss=0.3652, pruned_loss=0.1425, over 6503.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3335, pruned_loss=0.09963, over 1331483.86 frames. ], batch size: 71, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:26:30,598 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:26:43,233 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48874.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:27:00,372 INFO [train.py:901] (1/4) Epoch 7, batch 400, loss[loss=0.3472, simple_loss=0.3935, pruned_loss=0.1504, over 8707.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3357, pruned_loss=0.1011, over 1398708.16 frames. ], batch size: 34, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:27:00,663 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3763, 1.9433, 3.0639, 2.3834, 2.5771, 2.1118, 1.5573, 1.2673], + device='cuda:1'), covar=tensor([0.2643, 0.2907, 0.0653, 0.1717, 0.1353, 0.1439, 0.1294, 0.2977], + device='cuda:1'), in_proj_covar=tensor([0.0816, 0.0754, 0.0650, 0.0743, 0.0840, 0.0695, 0.0646, 0.0683], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:27:01,278 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48900.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:27:22,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.734e+02 3.619e+02 4.506e+02 1.679e+03, threshold=7.237e+02, percent-clipped=8.0 +2023-02-06 03:27:32,150 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48945.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:27:34,592 INFO [train.py:901] (1/4) Epoch 7, batch 450, loss[loss=0.2549, simple_loss=0.3103, pruned_loss=0.09975, over 7703.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3373, pruned_loss=0.1024, over 1446516.58 frames. ], batch size: 18, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:27:36,183 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5453, 1.9078, 2.1159, 1.1231, 2.2395, 1.3189, 0.5630, 1.6605], + device='cuda:1'), covar=tensor([0.0283, 0.0156, 0.0092, 0.0228, 0.0135, 0.0470, 0.0404, 0.0136], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0250, 0.0206, 0.0307, 0.0240, 0.0393, 0.0317, 0.0293], + device='cuda:1'), out_proj_covar=tensor([1.1364e-04, 7.9519e-05, 6.4796e-05, 9.7593e-05, 7.7148e-05, 1.3655e-04, + 1.0325e-04, 9.4133e-05], device='cuda:1') +2023-02-06 03:27:49,648 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48970.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:28:10,035 INFO [train.py:901] (1/4) Epoch 7, batch 500, loss[loss=0.3007, simple_loss=0.3685, pruned_loss=0.1165, over 8776.00 frames. ], tot_loss[loss=0.2712, simple_loss=0.3374, pruned_loss=0.1025, over 1481682.49 frames. ], batch size: 30, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:28:10,363 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 03:28:32,338 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.572e+02 3.184e+02 4.227e+02 8.649e+02, threshold=6.369e+02, percent-clipped=1.0 +2023-02-06 03:28:43,905 INFO [train.py:901] (1/4) Epoch 7, batch 550, loss[loss=0.2859, simple_loss=0.3474, pruned_loss=0.1121, over 8606.00 frames. ], tot_loss[loss=0.2707, simple_loss=0.3371, pruned_loss=0.1022, over 1509144.64 frames. ], batch size: 31, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:28:50,294 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:29:07,258 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:29:19,498 INFO [train.py:901] (1/4) Epoch 7, batch 600, loss[loss=0.2904, simple_loss=0.3477, pruned_loss=0.1165, over 7254.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3364, pruned_loss=0.1024, over 1527933.37 frames. ], batch size: 16, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:29:19,610 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0645, 3.9933, 3.7374, 2.0913, 3.5863, 3.5936, 3.7779, 3.2013], + device='cuda:1'), covar=tensor([0.0875, 0.0665, 0.0842, 0.4064, 0.0859, 0.1001, 0.1102, 0.0911], + device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0316, 0.0338, 0.0421, 0.0326, 0.0309, 0.0314, 0.0268], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:29:31,450 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 03:29:41,662 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49130.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:29:42,820 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.845e+02 3.510e+02 4.694e+02 1.227e+03, threshold=7.020e+02, percent-clipped=5.0 +2023-02-06 03:29:54,592 INFO [train.py:901] (1/4) Epoch 7, batch 650, loss[loss=0.2486, simple_loss=0.3231, pruned_loss=0.08701, over 8290.00 frames. ], tot_loss[loss=0.2708, simple_loss=0.337, pruned_loss=0.1023, over 1551935.08 frames. ], batch size: 23, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:29:55,667 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 +2023-02-06 03:29:58,991 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49155.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:29:59,738 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49156.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:30:17,689 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49181.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:30:18,993 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49183.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:30:24,598 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.9322, 1.0984, 4.2981, 1.6691, 3.1684, 3.2694, 3.7632, 3.7574], + device='cuda:1'), covar=tensor([0.1019, 0.6042, 0.0994, 0.4123, 0.2663, 0.1863, 0.1153, 0.1154], + device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0517, 0.0461, 0.0457, 0.0520, 0.0430, 0.0433, 0.0489], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 03:30:29,835 INFO [train.py:901] (1/4) Epoch 7, batch 700, loss[loss=0.2342, simple_loss=0.3079, pruned_loss=0.08022, over 8191.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3369, pruned_loss=0.1025, over 1565574.81 frames. ], batch size: 23, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:30:31,264 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49200.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:30:54,557 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.106e+02 2.876e+02 3.436e+02 4.276e+02 6.994e+02, threshold=6.873e+02, percent-clipped=0.0 +2023-02-06 03:31:06,154 INFO [train.py:901] (1/4) Epoch 7, batch 750, loss[loss=0.228, simple_loss=0.2978, pruned_loss=0.07904, over 8077.00 frames. ], tot_loss[loss=0.2694, simple_loss=0.3355, pruned_loss=0.1016, over 1578731.18 frames. ], batch size: 21, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:31:09,788 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49254.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:31:18,092 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 03:31:18,490 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.41 vs. limit=5.0 +2023-02-06 03:31:26,438 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 03:31:40,971 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49298.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:31:41,445 INFO [train.py:901] (1/4) Epoch 7, batch 800, loss[loss=0.2316, simple_loss=0.3009, pruned_loss=0.0811, over 8243.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3348, pruned_loss=0.1006, over 1592718.62 frames. ], batch size: 22, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:31:53,378 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49315.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:32:05,544 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.006e+02 2.832e+02 3.318e+02 4.162e+02 1.224e+03, threshold=6.636e+02, percent-clipped=6.0 +2023-02-06 03:32:17,948 INFO [train.py:901] (1/4) Epoch 7, batch 850, loss[loss=0.2286, simple_loss=0.2912, pruned_loss=0.08299, over 7437.00 frames. ], tot_loss[loss=0.2669, simple_loss=0.3338, pruned_loss=0.09995, over 1596095.21 frames. ], batch size: 17, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:32:52,561 INFO [train.py:901] (1/4) Epoch 7, batch 900, loss[loss=0.3344, simple_loss=0.3773, pruned_loss=0.1457, over 6754.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3351, pruned_loss=0.1009, over 1600040.75 frames. ], batch size: 71, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:33:07,535 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-06 03:33:10,011 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4595, 1.7695, 1.7483, 0.7402, 1.7364, 1.3549, 0.3701, 1.6251], + device='cuda:1'), covar=tensor([0.0213, 0.0123, 0.0104, 0.0235, 0.0157, 0.0380, 0.0375, 0.0099], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0254, 0.0210, 0.0308, 0.0242, 0.0395, 0.0317, 0.0294], + device='cuda:1'), out_proj_covar=tensor([1.1440e-04, 8.0497e-05, 6.5810e-05, 9.7596e-05, 7.7650e-05, 1.3680e-04, + 1.0305e-04, 9.4303e-05], device='cuda:1') +2023-02-06 03:33:14,970 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.14 vs. limit=5.0 +2023-02-06 03:33:17,136 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.812e+02 3.278e+02 4.578e+02 1.649e+03, threshold=6.556e+02, percent-clipped=8.0 +2023-02-06 03:33:22,011 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49440.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:33:28,088 INFO [train.py:901] (1/4) Epoch 7, batch 950, loss[loss=0.2527, simple_loss=0.3043, pruned_loss=0.1005, over 7534.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.3348, pruned_loss=0.101, over 1602521.33 frames. ], batch size: 18, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:33:28,882 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49450.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:33:50,525 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 03:34:03,669 INFO [train.py:901] (1/4) Epoch 7, batch 1000, loss[loss=0.2791, simple_loss=0.3462, pruned_loss=0.1059, over 8355.00 frames. ], tot_loss[loss=0.2698, simple_loss=0.3356, pruned_loss=0.102, over 1603558.28 frames. ], batch size: 24, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:34:18,333 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3033, 1.4914, 1.3129, 1.8804, 0.8896, 1.1176, 1.2682, 1.4635], + device='cuda:1'), covar=tensor([0.1074, 0.1021, 0.1416, 0.0634, 0.1251, 0.1952, 0.1084, 0.0936], + device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0245, 0.0283, 0.0227, 0.0243, 0.0274, 0.0277, 0.0244], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 03:34:24,221 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 03:34:27,700 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 3.091e+02 3.599e+02 4.515e+02 1.445e+03, threshold=7.198e+02, percent-clipped=7.0 +2023-02-06 03:34:35,939 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 03:34:38,748 INFO [train.py:901] (1/4) Epoch 7, batch 1050, loss[loss=0.2889, simple_loss=0.3611, pruned_loss=0.1084, over 8583.00 frames. ], tot_loss[loss=0.2696, simple_loss=0.3356, pruned_loss=0.1018, over 1608288.22 frames. ], batch size: 31, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:34:43,073 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49554.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:34:45,404 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.35 vs. limit=5.0 +2023-02-06 03:34:54,566 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49571.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:35:00,763 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49579.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:35:13,219 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2018, 1.5857, 3.3029, 1.2283, 2.2021, 3.6942, 3.6453, 3.1542], + device='cuda:1'), covar=tensor([0.0919, 0.1433, 0.0409, 0.2166, 0.0877, 0.0255, 0.0399, 0.0601], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0274, 0.0230, 0.0270, 0.0233, 0.0213, 0.0266, 0.0277], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 03:35:13,270 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49596.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:35:14,562 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49598.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:35:15,154 INFO [train.py:901] (1/4) Epoch 7, batch 1100, loss[loss=0.2836, simple_loss=0.3523, pruned_loss=0.1075, over 8600.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3362, pruned_loss=0.1025, over 1606808.18 frames. ], batch size: 31, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:35:38,352 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 2.766e+02 3.386e+02 4.310e+02 6.415e+02, threshold=6.771e+02, percent-clipped=0.0 +2023-02-06 03:35:46,089 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 03:35:49,456 INFO [train.py:901] (1/4) Epoch 7, batch 1150, loss[loss=0.2532, simple_loss=0.3279, pruned_loss=0.0892, over 8507.00 frames. ], tot_loss[loss=0.2707, simple_loss=0.3362, pruned_loss=0.1026, over 1610258.89 frames. ], batch size: 26, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:36:23,376 INFO [train.py:901] (1/4) Epoch 7, batch 1200, loss[loss=0.2906, simple_loss=0.3603, pruned_loss=0.1104, over 8756.00 frames. ], tot_loss[loss=0.2705, simple_loss=0.3361, pruned_loss=0.1024, over 1612654.67 frames. ], batch size: 30, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:36:33,697 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49713.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:36:41,591 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-02-06 03:36:47,173 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.006e+02 2.915e+02 3.820e+02 5.048e+02 1.193e+03, threshold=7.640e+02, percent-clipped=11.0 +2023-02-06 03:36:57,981 INFO [train.py:901] (1/4) Epoch 7, batch 1250, loss[loss=0.3111, simple_loss=0.3841, pruned_loss=0.119, over 8189.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3352, pruned_loss=0.1014, over 1610051.70 frames. ], batch size: 23, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:37:22,186 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49784.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:37:22,256 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49784.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:37:29,678 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49794.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:37:32,809 INFO [train.py:901] (1/4) Epoch 7, batch 1300, loss[loss=0.3144, simple_loss=0.3781, pruned_loss=0.1253, over 8578.00 frames. ], tot_loss[loss=0.2697, simple_loss=0.3363, pruned_loss=0.1015, over 1613659.05 frames. ], batch size: 39, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:37:57,661 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.968e+02 2.650e+02 3.390e+02 4.402e+02 9.600e+02, threshold=6.781e+02, percent-clipped=3.0 +2023-02-06 03:38:08,117 INFO [train.py:901] (1/4) Epoch 7, batch 1350, loss[loss=0.2852, simple_loss=0.3586, pruned_loss=0.1059, over 8317.00 frames. ], tot_loss[loss=0.2698, simple_loss=0.3365, pruned_loss=0.1016, over 1616110.99 frames. ], batch size: 26, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:38:42,144 INFO [train.py:901] (1/4) Epoch 7, batch 1400, loss[loss=0.2334, simple_loss=0.306, pruned_loss=0.08042, over 7784.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3367, pruned_loss=0.1023, over 1612990.41 frames. ], batch size: 19, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:38:43,020 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49899.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:38:49,796 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49909.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:39:07,159 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.921e+02 3.790e+02 4.996e+02 8.997e+02, threshold=7.579e+02, percent-clipped=6.0 +2023-02-06 03:39:11,375 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 03:39:12,310 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8607, 1.5324, 2.2235, 1.8423, 1.9697, 1.7076, 1.3712, 0.6241], + device='cuda:1'), covar=tensor([0.2724, 0.2561, 0.0767, 0.1433, 0.1231, 0.1431, 0.1330, 0.2717], + device='cuda:1'), in_proj_covar=tensor([0.0831, 0.0759, 0.0662, 0.0753, 0.0851, 0.0700, 0.0656, 0.0698], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:39:18,098 INFO [train.py:901] (1/4) Epoch 7, batch 1450, loss[loss=0.2549, simple_loss=0.3364, pruned_loss=0.08673, over 8754.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3364, pruned_loss=0.1017, over 1614263.47 frames. ], batch size: 30, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:39:31,680 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:39:49,082 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49994.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:39:52,398 INFO [train.py:901] (1/4) Epoch 7, batch 1500, loss[loss=0.2944, simple_loss=0.361, pruned_loss=0.1139, over 8347.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.337, pruned_loss=0.1024, over 1614182.37 frames. ], batch size: 26, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:40:16,582 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.823e+02 3.555e+02 4.038e+02 9.229e+02, threshold=7.110e+02, percent-clipped=3.0 +2023-02-06 03:40:27,896 INFO [train.py:901] (1/4) Epoch 7, batch 1550, loss[loss=0.3633, simple_loss=0.4005, pruned_loss=0.163, over 6913.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.3371, pruned_loss=0.1017, over 1616366.50 frames. ], batch size: 71, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:40:50,281 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7574, 1.1683, 3.9291, 1.4053, 3.4158, 3.2790, 3.5101, 3.3877], + device='cuda:1'), covar=tensor([0.0477, 0.3716, 0.0459, 0.2731, 0.1200, 0.0708, 0.0530, 0.0647], + device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0508, 0.0461, 0.0449, 0.0513, 0.0427, 0.0431, 0.0482], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 03:41:02,321 INFO [train.py:901] (1/4) Epoch 7, batch 1600, loss[loss=0.3494, simple_loss=0.3979, pruned_loss=0.1505, over 8110.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3363, pruned_loss=0.1009, over 1616993.32 frames. ], batch size: 23, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:41:22,664 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:41:25,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.849e+02 3.464e+02 4.418e+02 7.019e+02, threshold=6.928e+02, percent-clipped=0.0 +2023-02-06 03:41:35,387 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50146.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:41:37,165 INFO [train.py:901] (1/4) Epoch 7, batch 1650, loss[loss=0.3246, simple_loss=0.3707, pruned_loss=0.1393, over 7158.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3354, pruned_loss=0.1009, over 1615238.29 frames. ], batch size: 71, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:41:41,431 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50155.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:41:48,656 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50165.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:41:59,248 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50180.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:42:05,854 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50190.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:42:11,957 INFO [train.py:901] (1/4) Epoch 7, batch 1700, loss[loss=0.2603, simple_loss=0.3417, pruned_loss=0.08944, over 8595.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3359, pruned_loss=0.1007, over 1619850.39 frames. ], batch size: 34, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:42:20,323 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5194, 1.7855, 4.2064, 1.8738, 2.4732, 4.8645, 4.7238, 4.1366], + device='cuda:1'), covar=tensor([0.1020, 0.1471, 0.0320, 0.1996, 0.0923, 0.0191, 0.0297, 0.0629], + device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0281, 0.0235, 0.0276, 0.0241, 0.0218, 0.0276, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 03:42:22,985 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50215.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:42:35,242 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.886e+02 3.481e+02 4.608e+02 1.233e+03, threshold=6.962e+02, percent-clipped=3.0 +2023-02-06 03:42:42,128 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50243.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:42:46,679 INFO [train.py:901] (1/4) Epoch 7, batch 1750, loss[loss=0.2974, simple_loss=0.3645, pruned_loss=0.1151, over 8357.00 frames. ], tot_loss[loss=0.2701, simple_loss=0.3368, pruned_loss=0.1017, over 1616910.35 frames. ], batch size: 24, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:43:21,551 INFO [train.py:901] (1/4) Epoch 7, batch 1800, loss[loss=0.2666, simple_loss=0.3343, pruned_loss=0.09939, over 8138.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3372, pruned_loss=0.1023, over 1616331.18 frames. ], batch size: 22, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:43:23,896 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 03:43:44,798 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 2.922e+02 3.562e+02 4.379e+02 1.030e+03, threshold=7.125e+02, percent-clipped=4.0 +2023-02-06 03:43:56,169 INFO [train.py:901] (1/4) Epoch 7, batch 1850, loss[loss=0.2852, simple_loss=0.3568, pruned_loss=0.1068, over 8330.00 frames. ], tot_loss[loss=0.2707, simple_loss=0.3374, pruned_loss=0.102, over 1618454.61 frames. ], batch size: 26, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:44:21,516 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0223, 1.2732, 1.5677, 1.2844, 1.1227, 1.2854, 1.4725, 1.3685], + device='cuda:1'), covar=tensor([0.0561, 0.1354, 0.1785, 0.1451, 0.0631, 0.1735, 0.0787, 0.0621], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0169, 0.0209, 0.0172, 0.0117, 0.0177, 0.0130, 0.0148], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 03:44:30,595 INFO [train.py:901] (1/4) Epoch 7, batch 1900, loss[loss=0.2511, simple_loss=0.3217, pruned_loss=0.09026, over 8255.00 frames. ], tot_loss[loss=0.27, simple_loss=0.3368, pruned_loss=0.1016, over 1613172.55 frames. ], batch size: 22, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:44:43,985 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 03:44:53,923 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.786e+02 3.637e+02 4.614e+02 8.948e+02, threshold=7.273e+02, percent-clipped=3.0 +2023-02-06 03:44:56,011 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 03:45:04,117 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6825, 1.8596, 2.1138, 1.7160, 1.0055, 2.1065, 0.4872, 1.3100], + device='cuda:1'), covar=tensor([0.3109, 0.1641, 0.0850, 0.2305, 0.5851, 0.0654, 0.4396, 0.2500], + device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0136, 0.0082, 0.0188, 0.0227, 0.0086, 0.0145, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:45:04,598 INFO [train.py:901] (1/4) Epoch 7, batch 1950, loss[loss=0.2773, simple_loss=0.3501, pruned_loss=0.1023, over 8556.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.3354, pruned_loss=0.1007, over 1614183.80 frames. ], batch size: 31, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:45:15,317 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 03:45:27,055 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-02-06 03:45:33,290 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:45:39,206 INFO [train.py:901] (1/4) Epoch 7, batch 2000, loss[loss=0.281, simple_loss=0.3522, pruned_loss=0.1049, over 8507.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3324, pruned_loss=0.09936, over 1606302.28 frames. ], batch size: 28, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:45:39,413 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50499.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:45:47,633 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-06 03:45:56,765 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:46:03,218 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.800e+02 3.583e+02 4.591e+02 1.075e+03, threshold=7.166e+02, percent-clipped=7.0 +2023-02-06 03:46:13,437 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9552, 1.4776, 1.5248, 1.3560, 1.0524, 1.3483, 1.5569, 1.5057], + device='cuda:1'), covar=tensor([0.0559, 0.1211, 0.1662, 0.1368, 0.0634, 0.1553, 0.0777, 0.0585], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0169, 0.0208, 0.0171, 0.0118, 0.0176, 0.0130, 0.0147], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 03:46:13,940 INFO [train.py:901] (1/4) Epoch 7, batch 2050, loss[loss=0.2555, simple_loss=0.33, pruned_loss=0.09052, over 8470.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3312, pruned_loss=0.09836, over 1604230.97 frames. ], batch size: 29, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:46:20,542 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50559.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:46:23,779 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50564.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:46:47,288 INFO [train.py:901] (1/4) Epoch 7, batch 2100, loss[loss=0.3133, simple_loss=0.3672, pruned_loss=0.1297, over 8106.00 frames. ], tot_loss[loss=0.2666, simple_loss=0.3332, pruned_loss=0.1, over 1605480.06 frames. ], batch size: 23, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:46:52,374 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:47:11,024 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 3.066e+02 3.697e+02 4.610e+02 1.063e+03, threshold=7.394e+02, percent-clipped=3.0 +2023-02-06 03:47:12,591 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4538, 2.0764, 3.5686, 1.2167, 2.5094, 1.9604, 1.7291, 2.0940], + device='cuda:1'), covar=tensor([0.1572, 0.1774, 0.0637, 0.3197, 0.1445, 0.2389, 0.1492, 0.2398], + device='cuda:1'), in_proj_covar=tensor([0.0473, 0.0465, 0.0528, 0.0547, 0.0593, 0.0528, 0.0450, 0.0590], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 03:47:22,368 INFO [train.py:901] (1/4) Epoch 7, batch 2150, loss[loss=0.2615, simple_loss=0.3363, pruned_loss=0.09335, over 8560.00 frames. ], tot_loss[loss=0.2666, simple_loss=0.3331, pruned_loss=0.1001, over 1606743.72 frames. ], batch size: 31, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:47:40,256 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50674.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:47:44,608 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 03:47:52,461 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.38 vs. limit=5.0 +2023-02-06 03:47:56,609 INFO [train.py:901] (1/4) Epoch 7, batch 2200, loss[loss=0.2915, simple_loss=0.3677, pruned_loss=0.1077, over 8588.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3337, pruned_loss=0.1001, over 1605562.53 frames. ], batch size: 31, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:48:09,854 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.92 vs. limit=5.0 +2023-02-06 03:48:20,831 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.939e+02 3.492e+02 4.230e+02 8.261e+02, threshold=6.983e+02, percent-clipped=2.0 +2023-02-06 03:48:31,234 INFO [train.py:901] (1/4) Epoch 7, batch 2250, loss[loss=0.2497, simple_loss=0.3163, pruned_loss=0.09161, over 7799.00 frames. ], tot_loss[loss=0.2662, simple_loss=0.3333, pruned_loss=0.09959, over 1606334.30 frames. ], batch size: 20, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:49:04,149 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6236, 1.2497, 2.7657, 1.1258, 1.9459, 3.0014, 3.0451, 2.5463], + device='cuda:1'), covar=tensor([0.0952, 0.1482, 0.0436, 0.2138, 0.0796, 0.0308, 0.0474, 0.0695], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0270, 0.0227, 0.0267, 0.0236, 0.0212, 0.0266, 0.0278], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 03:49:05,375 INFO [train.py:901] (1/4) Epoch 7, batch 2300, loss[loss=0.2661, simple_loss=0.3443, pruned_loss=0.09398, over 8471.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3347, pruned_loss=0.1006, over 1608142.30 frames. ], batch size: 25, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:49:06,293 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50800.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:49:23,393 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50826.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:49:28,611 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 3.140e+02 4.073e+02 5.620e+02 1.608e+03, threshold=8.146e+02, percent-clipped=16.0 +2023-02-06 03:49:39,827 INFO [train.py:901] (1/4) Epoch 7, batch 2350, loss[loss=0.2812, simple_loss=0.3548, pruned_loss=0.1039, over 8462.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3358, pruned_loss=0.1012, over 1613881.36 frames. ], batch size: 27, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:49:47,977 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50861.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:50:04,997 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:50:14,216 INFO [train.py:901] (1/4) Epoch 7, batch 2400, loss[loss=0.267, simple_loss=0.3471, pruned_loss=0.0934, over 8246.00 frames. ], tot_loss[loss=0.268, simple_loss=0.335, pruned_loss=0.1005, over 1614673.03 frames. ], batch size: 24, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:50:20,197 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50908.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:50:30,576 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-02-06 03:50:35,196 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50930.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:50:36,987 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.859e+02 2.900e+02 3.414e+02 4.196e+02 7.276e+02, threshold=6.828e+02, percent-clipped=0.0 +2023-02-06 03:50:47,541 INFO [train.py:901] (1/4) Epoch 7, batch 2450, loss[loss=0.2652, simple_loss=0.3347, pruned_loss=0.09786, over 8340.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3348, pruned_loss=0.1006, over 1613582.69 frames. ], batch size: 49, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:50:52,350 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50955.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:50:59,075 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50964.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:51:22,800 INFO [train.py:901] (1/4) Epoch 7, batch 2500, loss[loss=0.2613, simple_loss=0.3399, pruned_loss=0.09131, over 8461.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3347, pruned_loss=0.1004, over 1610474.21 frames. ], batch size: 27, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:51:36,588 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3006, 1.7961, 4.4685, 2.0250, 3.8978, 3.7675, 4.0843, 3.8912], + device='cuda:1'), covar=tensor([0.0531, 0.3650, 0.0441, 0.2685, 0.1052, 0.0799, 0.0553, 0.0621], + device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0526, 0.0475, 0.0456, 0.0528, 0.0435, 0.0437, 0.0497], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 03:51:39,925 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51023.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:51:45,169 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3255, 1.4600, 4.4549, 1.4885, 3.8728, 3.6972, 3.9919, 3.8156], + device='cuda:1'), covar=tensor([0.0450, 0.3845, 0.0367, 0.2998, 0.0970, 0.0713, 0.0521, 0.0624], + device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0524, 0.0472, 0.0456, 0.0524, 0.0433, 0.0435, 0.0493], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 03:51:46,387 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.634e+02 3.421e+02 4.023e+02 8.503e+02, threshold=6.842e+02, percent-clipped=1.0 +2023-02-06 03:51:56,889 INFO [train.py:901] (1/4) Epoch 7, batch 2550, loss[loss=0.2532, simple_loss=0.3374, pruned_loss=0.08444, over 8555.00 frames. ], tot_loss[loss=0.2683, simple_loss=0.3355, pruned_loss=0.1006, over 1615030.82 frames. ], batch size: 31, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:52:18,133 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:52:31,079 INFO [train.py:901] (1/4) Epoch 7, batch 2600, loss[loss=0.3835, simple_loss=0.4033, pruned_loss=0.1819, over 6449.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.3348, pruned_loss=0.1003, over 1610630.72 frames. ], batch size: 71, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:52:54,868 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.187e+02 3.140e+02 3.874e+02 4.757e+02 8.436e+02, threshold=7.747e+02, percent-clipped=5.0 +2023-02-06 03:53:02,023 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51144.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:53:05,174 INFO [train.py:901] (1/4) Epoch 7, batch 2650, loss[loss=0.3667, simple_loss=0.4029, pruned_loss=0.1653, over 8461.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3353, pruned_loss=0.1003, over 1610647.11 frames. ], batch size: 25, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:53:19,254 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51170.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:53:32,643 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51190.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:53:39,227 INFO [train.py:901] (1/4) Epoch 7, batch 2700, loss[loss=0.2885, simple_loss=0.3599, pruned_loss=0.1086, over 8487.00 frames. ], tot_loss[loss=0.2675, simple_loss=0.335, pruned_loss=0.09998, over 1608468.13 frames. ], batch size: 39, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:53:58,088 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2493, 1.4741, 2.1771, 1.1395, 1.5344, 1.5534, 1.3254, 1.3853], + device='cuda:1'), covar=tensor([0.1678, 0.1860, 0.0751, 0.3316, 0.1345, 0.2506, 0.1686, 0.1776], + device='cuda:1'), in_proj_covar=tensor([0.0475, 0.0471, 0.0540, 0.0553, 0.0591, 0.0534, 0.0453, 0.0596], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 03:54:02,467 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.830e+02 3.046e+02 3.584e+02 4.560e+02 9.753e+02, threshold=7.169e+02, percent-clipped=4.0 +2023-02-06 03:54:14,246 INFO [train.py:901] (1/4) Epoch 7, batch 2750, loss[loss=0.2639, simple_loss=0.3178, pruned_loss=0.105, over 7935.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3356, pruned_loss=0.1002, over 1614448.28 frames. ], batch size: 20, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:54:21,062 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51259.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:54:34,381 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51279.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:54:38,289 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51285.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:54:47,538 INFO [train.py:901] (1/4) Epoch 7, batch 2800, loss[loss=0.2447, simple_loss=0.31, pruned_loss=0.08972, over 7689.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.335, pruned_loss=0.09969, over 1618740.43 frames. ], batch size: 18, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:54:51,138 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51304.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:54:53,735 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51308.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:55:11,784 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.811e+02 3.563e+02 4.674e+02 6.809e+02, threshold=7.126e+02, percent-clipped=0.0 +2023-02-06 03:55:14,047 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0070, 2.0784, 1.8379, 2.6545, 1.0102, 1.5606, 1.6553, 2.0609], + device='cuda:1'), covar=tensor([0.0799, 0.1011, 0.1334, 0.0544, 0.1640, 0.1765, 0.1316, 0.0959], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0239, 0.0278, 0.0222, 0.0243, 0.0267, 0.0278, 0.0243], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 03:55:22,690 INFO [train.py:901] (1/4) Epoch 7, batch 2850, loss[loss=0.219, simple_loss=0.2936, pruned_loss=0.07217, over 7550.00 frames. ], tot_loss[loss=0.2644, simple_loss=0.3326, pruned_loss=0.09812, over 1611399.10 frames. ], batch size: 18, lr: 1.10e-02, grad_scale: 16.0 +2023-02-06 03:55:57,259 INFO [train.py:901] (1/4) Epoch 7, batch 2900, loss[loss=0.2776, simple_loss=0.3358, pruned_loss=0.1097, over 8265.00 frames. ], tot_loss[loss=0.2649, simple_loss=0.3324, pruned_loss=0.09875, over 1608252.06 frames. ], batch size: 24, lr: 1.10e-02, grad_scale: 16.0 +2023-02-06 03:55:58,210 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51400.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:56:04,948 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51410.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:56:13,544 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51423.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:56:14,126 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51424.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:56:19,622 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 03:56:20,271 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 2.833e+02 3.577e+02 4.732e+02 1.075e+03, threshold=7.153e+02, percent-clipped=9.0 +2023-02-06 03:56:32,326 INFO [train.py:901] (1/4) Epoch 7, batch 2950, loss[loss=0.2751, simple_loss=0.3388, pruned_loss=0.1057, over 8243.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3325, pruned_loss=0.09945, over 1606013.90 frames. ], batch size: 22, lr: 1.10e-02, grad_scale: 16.0 +2023-02-06 03:56:34,164 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.87 vs. limit=5.0 +2023-02-06 03:56:38,234 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-06 03:57:06,418 INFO [train.py:901] (1/4) Epoch 7, batch 3000, loss[loss=0.28, simple_loss=0.343, pruned_loss=0.1084, over 8516.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.333, pruned_loss=0.09998, over 1608440.89 frames. ], batch size: 28, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:57:06,419 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 03:57:21,704 INFO [train.py:935] (1/4) Epoch 7, validation: loss=0.2071, simple_loss=0.305, pruned_loss=0.05459, over 944034.00 frames. +2023-02-06 03:57:21,706 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6607MB +2023-02-06 03:57:31,203 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51513.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:57:32,577 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51515.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:57:37,987 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7464, 2.3141, 3.8596, 2.8168, 3.0870, 2.2540, 1.6774, 1.7278], + device='cuda:1'), covar=tensor([0.2546, 0.3101, 0.0642, 0.1694, 0.1396, 0.1445, 0.1336, 0.3309], + device='cuda:1'), in_proj_covar=tensor([0.0817, 0.0756, 0.0656, 0.0750, 0.0845, 0.0694, 0.0653, 0.0692], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:57:45,152 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.842e+02 3.422e+02 4.197e+02 1.269e+03, threshold=6.844e+02, percent-clipped=2.0 +2023-02-06 03:57:45,245 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51534.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:57:48,650 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51539.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:57:49,353 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51540.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:57:50,068 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51541.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:57:55,343 INFO [train.py:901] (1/4) Epoch 7, batch 3050, loss[loss=0.2866, simple_loss=0.3566, pruned_loss=0.1083, over 8296.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.333, pruned_loss=0.09934, over 1609693.11 frames. ], batch size: 23, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:58:06,943 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51566.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:58:12,896 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0280, 1.5890, 1.6093, 1.3060, 1.0742, 1.3904, 1.6936, 1.5018], + device='cuda:1'), covar=tensor([0.0543, 0.1197, 0.1664, 0.1350, 0.0595, 0.1465, 0.0681, 0.0560], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0167, 0.0210, 0.0173, 0.0117, 0.0174, 0.0128, 0.0147], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 03:58:29,887 INFO [train.py:901] (1/4) Epoch 7, batch 3100, loss[loss=0.2509, simple_loss=0.309, pruned_loss=0.09637, over 7253.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.334, pruned_loss=0.1002, over 1612929.53 frames. ], batch size: 16, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:58:54,840 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.934e+02 3.035e+02 3.902e+02 5.145e+02 1.067e+03, threshold=7.804e+02, percent-clipped=7.0 +2023-02-06 03:59:05,332 INFO [train.py:901] (1/4) Epoch 7, batch 3150, loss[loss=0.2636, simple_loss=0.3334, pruned_loss=0.0969, over 8612.00 frames. ], tot_loss[loss=0.2682, simple_loss=0.3347, pruned_loss=0.1009, over 1609958.59 frames. ], batch size: 34, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:59:05,517 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51649.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:59:23,453 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6366, 4.7051, 4.0884, 1.8332, 4.0207, 4.1275, 4.3128, 3.6289], + device='cuda:1'), covar=tensor([0.0802, 0.0597, 0.1025, 0.4871, 0.0843, 0.0747, 0.1364, 0.0963], + device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0320, 0.0346, 0.0429, 0.0334, 0.0313, 0.0325, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:59:26,411 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51679.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:59:33,466 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8702, 2.3299, 3.9370, 2.8586, 3.1407, 2.3181, 1.8178, 1.8206], + device='cuda:1'), covar=tensor([0.2360, 0.2909, 0.0645, 0.1693, 0.1469, 0.1465, 0.1340, 0.3203], + device='cuda:1'), in_proj_covar=tensor([0.0821, 0.0758, 0.0666, 0.0753, 0.0854, 0.0706, 0.0655, 0.0698], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 03:59:40,113 INFO [train.py:901] (1/4) Epoch 7, batch 3200, loss[loss=0.3204, simple_loss=0.3774, pruned_loss=0.1317, over 8547.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.3345, pruned_loss=0.1004, over 1611925.39 frames. ], batch size: 34, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:59:43,600 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51704.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:05,271 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 2.946e+02 3.588e+02 4.680e+02 7.788e+02, threshold=7.176e+02, percent-clipped=0.0 +2023-02-06 04:00:12,048 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51744.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:16,041 INFO [train.py:901] (1/4) Epoch 7, batch 3250, loss[loss=0.2179, simple_loss=0.2931, pruned_loss=0.07136, over 8233.00 frames. ], tot_loss[loss=0.2676, simple_loss=0.3345, pruned_loss=0.1004, over 1610818.94 frames. ], batch size: 22, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:00:19,464 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:23,521 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8239, 1.5968, 1.6981, 1.3853, 1.3797, 1.5896, 2.1727, 1.9054], + device='cuda:1'), covar=tensor([0.0551, 0.1330, 0.1838, 0.1523, 0.0659, 0.1613, 0.0704, 0.0650], + device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0166, 0.0210, 0.0172, 0.0117, 0.0174, 0.0127, 0.0147], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 04:00:26,784 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51765.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:46,801 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51795.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:49,339 INFO [train.py:901] (1/4) Epoch 7, batch 3300, loss[loss=0.2353, simple_loss=0.3035, pruned_loss=0.08352, over 7812.00 frames. ], tot_loss[loss=0.2669, simple_loss=0.3336, pruned_loss=0.1001, over 1611207.41 frames. ], batch size: 20, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:01:03,665 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51820.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:01:14,384 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.988e+02 3.662e+02 4.246e+02 9.313e+02, threshold=7.324e+02, percent-clipped=2.0 +2023-02-06 04:01:24,626 INFO [train.py:901] (1/4) Epoch 7, batch 3350, loss[loss=0.2404, simple_loss=0.321, pruned_loss=0.07987, over 8102.00 frames. ], tot_loss[loss=0.266, simple_loss=0.3328, pruned_loss=0.09958, over 1611154.56 frames. ], batch size: 23, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:01:30,919 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51857.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:01:32,345 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51859.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:01:39,536 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51869.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:01:59,352 INFO [train.py:901] (1/4) Epoch 7, batch 3400, loss[loss=0.2775, simple_loss=0.3262, pruned_loss=0.1144, over 8131.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.3315, pruned_loss=0.09908, over 1609533.44 frames. ], batch size: 22, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:02:03,642 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51905.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:02:10,928 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-06 04:02:16,129 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0763, 2.2641, 1.7812, 2.6311, 1.4914, 1.4978, 1.9555, 2.3020], + device='cuda:1'), covar=tensor([0.0785, 0.0851, 0.1138, 0.0508, 0.1205, 0.1573, 0.1112, 0.0764], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0233, 0.0275, 0.0219, 0.0239, 0.0263, 0.0272, 0.0237], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 04:02:20,935 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51930.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:02:23,259 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 2.723e+02 3.470e+02 4.144e+02 7.359e+02, threshold=6.940e+02, percent-clipped=1.0 +2023-02-06 04:02:24,205 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7168, 1.5062, 2.0072, 1.6962, 1.7716, 1.6425, 1.3076, 0.7041], + device='cuda:1'), covar=tensor([0.2443, 0.2354, 0.0709, 0.1264, 0.1129, 0.1357, 0.1247, 0.2290], + device='cuda:1'), in_proj_covar=tensor([0.0822, 0.0758, 0.0656, 0.0752, 0.0856, 0.0704, 0.0653, 0.0691], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:02:34,621 INFO [train.py:901] (1/4) Epoch 7, batch 3450, loss[loss=0.2552, simple_loss=0.3318, pruned_loss=0.08931, over 8482.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3337, pruned_loss=0.1004, over 1612454.54 frames. ], batch size: 27, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:02:50,955 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51972.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:03:04,798 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5337, 1.5771, 2.7559, 1.1015, 2.0850, 3.0586, 3.0385, 2.5618], + device='cuda:1'), covar=tensor([0.1100, 0.1284, 0.0450, 0.2190, 0.0715, 0.0307, 0.0512, 0.0742], + device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0274, 0.0230, 0.0271, 0.0243, 0.0218, 0.0275, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 04:03:09,380 INFO [train.py:901] (1/4) Epoch 7, batch 3500, loss[loss=0.2624, simple_loss=0.3452, pruned_loss=0.08976, over 8315.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.334, pruned_loss=0.1001, over 1611456.17 frames. ], batch size: 25, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:03:22,435 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 04:03:27,422 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.58 vs. limit=5.0 +2023-02-06 04:03:33,496 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.822e+02 3.302e+02 4.435e+02 1.594e+03, threshold=6.604e+02, percent-clipped=5.0 +2023-02-06 04:03:42,067 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 04:03:43,734 INFO [train.py:901] (1/4) Epoch 7, batch 3550, loss[loss=0.2451, simple_loss=0.315, pruned_loss=0.08755, over 7969.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3338, pruned_loss=0.09976, over 1613612.07 frames. ], batch size: 21, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:03:48,709 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7709, 2.1391, 4.3148, 1.2792, 2.7149, 2.1457, 1.6679, 2.5480], + device='cuda:1'), covar=tensor([0.1577, 0.2247, 0.0690, 0.3625, 0.1654, 0.2691, 0.1736, 0.2492], + device='cuda:1'), in_proj_covar=tensor([0.0472, 0.0472, 0.0527, 0.0549, 0.0583, 0.0522, 0.0452, 0.0590], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 04:03:50,008 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52058.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:19,844 INFO [train.py:901] (1/4) Epoch 7, batch 3600, loss[loss=0.2119, simple_loss=0.2939, pruned_loss=0.06496, over 7542.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.333, pruned_loss=0.09875, over 1614658.89 frames. ], batch size: 18, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:04:24,802 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1161, 1.6859, 1.4000, 1.7892, 1.3451, 1.2075, 1.1654, 1.5390], + device='cuda:1'), covar=tensor([0.0844, 0.0359, 0.0914, 0.0351, 0.0638, 0.1062, 0.0703, 0.0582], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0237, 0.0316, 0.0303, 0.0316, 0.0322, 0.0347, 0.0325], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 04:04:26,702 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52109.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:28,341 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 04:04:30,885 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52115.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:37,712 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:43,468 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.818e+02 3.176e+02 4.094e+02 8.086e+02, threshold=6.353e+02, percent-clipped=5.0 +2023-02-06 04:04:47,733 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:53,395 INFO [train.py:901] (1/4) Epoch 7, batch 3650, loss[loss=0.3627, simple_loss=0.4009, pruned_loss=0.1623, over 6948.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3328, pruned_loss=0.09872, over 1613839.74 frames. ], batch size: 72, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:04:54,173 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52150.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:05:21,245 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.3024, 1.2828, 5.6125, 2.2657, 4.3314, 4.4292, 5.1407, 5.1194], + device='cuda:1'), covar=tensor([0.0849, 0.5849, 0.0651, 0.3272, 0.2116, 0.1186, 0.0730, 0.0723], + device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0526, 0.0475, 0.0468, 0.0528, 0.0436, 0.0429, 0.0493], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 04:05:23,189 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 04:05:28,581 INFO [train.py:901] (1/4) Epoch 7, batch 3700, loss[loss=0.2861, simple_loss=0.3547, pruned_loss=0.1087, over 8558.00 frames. ], tot_loss[loss=0.2663, simple_loss=0.3341, pruned_loss=0.09921, over 1613983.74 frames. ], batch size: 39, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:05:36,879 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52211.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:05:47,064 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52224.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:05:49,831 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52228.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:05:53,723 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.627e+02 2.602e+02 3.554e+02 4.404e+02 9.700e+02, threshold=7.108e+02, percent-clipped=5.0 +2023-02-06 04:06:04,133 INFO [train.py:901] (1/4) Epoch 7, batch 3750, loss[loss=0.2274, simple_loss=0.3048, pruned_loss=0.07494, over 8245.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3343, pruned_loss=0.09931, over 1615687.54 frames. ], batch size: 22, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:06:07,173 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52253.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:06:12,885 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7717, 2.0107, 1.6962, 2.5268, 1.2344, 1.4474, 1.8013, 1.9596], + device='cuda:1'), covar=tensor([0.0893, 0.1107, 0.1115, 0.0510, 0.1248, 0.1456, 0.1017, 0.0835], + device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0238, 0.0280, 0.0223, 0.0241, 0.0266, 0.0274, 0.0241], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 04:06:24,074 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.51 vs. limit=5.0 +2023-02-06 04:06:38,742 INFO [train.py:901] (1/4) Epoch 7, batch 3800, loss[loss=0.2463, simple_loss=0.3239, pruned_loss=0.0843, over 8297.00 frames. ], tot_loss[loss=0.266, simple_loss=0.3331, pruned_loss=0.09948, over 1606044.85 frames. ], batch size: 23, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:07:01,885 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52330.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:07:04,421 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.780e+02 3.361e+02 4.228e+02 6.516e+02, threshold=6.722e+02, percent-clipped=0.0 +2023-02-06 04:07:15,844 INFO [train.py:901] (1/4) Epoch 7, batch 3850, loss[loss=0.2396, simple_loss=0.3107, pruned_loss=0.08423, over 7973.00 frames. ], tot_loss[loss=0.2654, simple_loss=0.3331, pruned_loss=0.09885, over 1610550.48 frames. ], batch size: 21, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:07:30,480 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 04:07:49,738 INFO [train.py:901] (1/4) Epoch 7, batch 3900, loss[loss=0.2355, simple_loss=0.301, pruned_loss=0.08502, over 7788.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3335, pruned_loss=0.09916, over 1609036.15 frames. ], batch size: 19, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:07:51,996 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52402.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:08:15,069 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.697e+02 3.207e+02 4.225e+02 1.297e+03, threshold=6.414e+02, percent-clipped=5.0 +2023-02-06 04:08:25,215 INFO [train.py:901] (1/4) Epoch 7, batch 3950, loss[loss=0.2484, simple_loss=0.3239, pruned_loss=0.08644, over 8316.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3316, pruned_loss=0.09782, over 1612841.04 frames. ], batch size: 25, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:08:47,998 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52480.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:08:57,501 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0437, 1.0081, 4.1963, 1.5874, 3.6478, 3.4849, 3.7995, 3.6472], + device='cuda:1'), covar=tensor([0.0459, 0.4108, 0.0408, 0.2917, 0.1123, 0.0803, 0.0498, 0.0595], + device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0527, 0.0479, 0.0469, 0.0533, 0.0443, 0.0433, 0.0496], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 04:09:00,793 INFO [train.py:901] (1/4) Epoch 7, batch 4000, loss[loss=0.25, simple_loss=0.3081, pruned_loss=0.0959, over 7694.00 frames. ], tot_loss[loss=0.2635, simple_loss=0.3312, pruned_loss=0.09793, over 1609308.78 frames. ], batch size: 18, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:09:04,720 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.75 vs. limit=5.0 +2023-02-06 04:09:05,167 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52505.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:09:13,173 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52517.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:09:24,184 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 2.935e+02 3.629e+02 4.693e+02 1.248e+03, threshold=7.258e+02, percent-clipped=9.0 +2023-02-06 04:09:35,733 INFO [train.py:901] (1/4) Epoch 7, batch 4050, loss[loss=0.2372, simple_loss=0.3146, pruned_loss=0.0799, over 8198.00 frames. ], tot_loss[loss=0.2642, simple_loss=0.3316, pruned_loss=0.09838, over 1609546.30 frames. ], batch size: 23, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:09:39,788 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52555.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:09:52,444 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52573.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:10:10,474 INFO [train.py:901] (1/4) Epoch 7, batch 4100, loss[loss=0.2613, simple_loss=0.3328, pruned_loss=0.09491, over 8469.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3323, pruned_loss=0.09883, over 1612930.90 frames. ], batch size: 25, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:10:33,846 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.709e+02 3.346e+02 4.687e+02 1.096e+03, threshold=6.691e+02, percent-clipped=5.0 +2023-02-06 04:10:44,015 INFO [train.py:901] (1/4) Epoch 7, batch 4150, loss[loss=0.2652, simple_loss=0.348, pruned_loss=0.09116, over 8477.00 frames. ], tot_loss[loss=0.2644, simple_loss=0.3321, pruned_loss=0.09833, over 1615222.00 frames. ], batch size: 29, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:10:59,827 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52670.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:11:02,381 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52674.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:11:20,451 INFO [train.py:901] (1/4) Epoch 7, batch 4200, loss[loss=0.375, simple_loss=0.416, pruned_loss=0.167, over 8342.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3327, pruned_loss=0.09863, over 1616215.84 frames. ], batch size: 26, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:11:30,522 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 04:11:43,886 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.675e+02 3.334e+02 4.108e+02 1.082e+03, threshold=6.669e+02, percent-clipped=4.0 +2023-02-06 04:11:53,187 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 04:11:53,854 INFO [train.py:901] (1/4) Epoch 7, batch 4250, loss[loss=0.2295, simple_loss=0.2968, pruned_loss=0.08115, over 7551.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3328, pruned_loss=0.09835, over 1617222.21 frames. ], batch size: 18, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:12:05,291 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8664, 1.6880, 2.5382, 1.6291, 2.1111, 2.8032, 2.6662, 2.5741], + device='cuda:1'), covar=tensor([0.0702, 0.1037, 0.0623, 0.1437, 0.0939, 0.0288, 0.0605, 0.0476], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0274, 0.0229, 0.0271, 0.0241, 0.0216, 0.0274, 0.0278], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 04:12:08,145 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52770.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:12:10,304 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52773.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:12:22,419 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:12:28,472 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52798.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:12:28,947 INFO [train.py:901] (1/4) Epoch 7, batch 4300, loss[loss=0.2401, simple_loss=0.3177, pruned_loss=0.08128, over 8495.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3313, pruned_loss=0.09794, over 1612705.99 frames. ], batch size: 29, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:12:53,670 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.879e+02 3.462e+02 4.347e+02 1.112e+03, threshold=6.924e+02, percent-clipped=5.0 +2023-02-06 04:13:03,883 INFO [train.py:901] (1/4) Epoch 7, batch 4350, loss[loss=0.2297, simple_loss=0.3048, pruned_loss=0.07729, over 8108.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3306, pruned_loss=0.09714, over 1615000.54 frames. ], batch size: 23, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:13:24,436 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 04:13:38,502 INFO [train.py:901] (1/4) Epoch 7, batch 4400, loss[loss=0.2119, simple_loss=0.2877, pruned_loss=0.06804, over 7928.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3306, pruned_loss=0.09687, over 1618422.51 frames. ], batch size: 20, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:13:50,775 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52917.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:13:56,757 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52926.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:14:02,505 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.715e+02 3.689e+02 4.508e+02 8.331e+02, threshold=7.379e+02, percent-clipped=6.0 +2023-02-06 04:14:06,680 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 04:14:13,199 INFO [train.py:901] (1/4) Epoch 7, batch 4450, loss[loss=0.3267, simple_loss=0.3646, pruned_loss=0.1444, over 7794.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3297, pruned_loss=0.09646, over 1616148.85 frames. ], batch size: 19, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:14:14,720 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:14:19,438 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2355, 1.7768, 2.7316, 2.1572, 2.2278, 1.8956, 1.5695, 0.9851], + device='cuda:1'), covar=tensor([0.2431, 0.2573, 0.0633, 0.1358, 0.1238, 0.1424, 0.1287, 0.2702], + device='cuda:1'), in_proj_covar=tensor([0.0825, 0.0770, 0.0674, 0.0762, 0.0857, 0.0708, 0.0658, 0.0708], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:14:46,592 INFO [train.py:901] (1/4) Epoch 7, batch 4500, loss[loss=0.2082, simple_loss=0.2831, pruned_loss=0.0666, over 7543.00 frames. ], tot_loss[loss=0.2623, simple_loss=0.3309, pruned_loss=0.09684, over 1618102.98 frames. ], batch size: 18, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:14:49,877 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-06 04:14:59,362 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 04:15:10,359 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53032.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:15:11,037 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8204, 1.9300, 2.1852, 1.6707, 1.1148, 2.2836, 0.4322, 1.4787], + device='cuda:1'), covar=tensor([0.3475, 0.1620, 0.0653, 0.2531, 0.5420, 0.0717, 0.4255, 0.2007], + device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0143, 0.0086, 0.0193, 0.0233, 0.0089, 0.0150, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:15:11,488 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.236e+02 2.890e+02 3.405e+02 4.030e+02 1.067e+03, threshold=6.809e+02, percent-clipped=4.0 +2023-02-06 04:15:18,996 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:15:22,053 INFO [train.py:901] (1/4) Epoch 7, batch 4550, loss[loss=0.3094, simple_loss=0.3741, pruned_loss=0.1223, over 8599.00 frames. ], tot_loss[loss=0.2624, simple_loss=0.3309, pruned_loss=0.09697, over 1616603.49 frames. ], batch size: 31, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:15:36,992 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53070.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:15:39,629 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53074.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:15:56,219 INFO [train.py:901] (1/4) Epoch 7, batch 4600, loss[loss=0.23, simple_loss=0.3078, pruned_loss=0.07617, over 8084.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3307, pruned_loss=0.09674, over 1617553.99 frames. ], batch size: 21, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:16:06,504 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53114.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:16:20,975 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.951e+02 3.579e+02 4.375e+02 1.013e+03, threshold=7.158e+02, percent-clipped=5.0 +2023-02-06 04:16:31,857 INFO [train.py:901] (1/4) Epoch 7, batch 4650, loss[loss=0.2647, simple_loss=0.3364, pruned_loss=0.09653, over 7964.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3302, pruned_loss=0.09685, over 1615007.59 frames. ], batch size: 21, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:17:06,943 INFO [train.py:901] (1/4) Epoch 7, batch 4700, loss[loss=0.218, simple_loss=0.2986, pruned_loss=0.06871, over 7927.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.331, pruned_loss=0.09784, over 1608394.35 frames. ], batch size: 20, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:17:27,463 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53229.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:17:30,529 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.793e+02 3.469e+02 4.275e+02 9.300e+02, threshold=6.939e+02, percent-clipped=3.0 +2023-02-06 04:17:34,871 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6610, 2.3293, 4.7049, 1.3090, 3.1543, 2.0951, 1.7251, 2.9211], + device='cuda:1'), covar=tensor([0.1619, 0.1881, 0.0573, 0.3714, 0.1358, 0.2588, 0.1721, 0.2117], + device='cuda:1'), in_proj_covar=tensor([0.0475, 0.0474, 0.0527, 0.0551, 0.0597, 0.0534, 0.0455, 0.0588], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:17:41,199 INFO [train.py:901] (1/4) Epoch 7, batch 4750, loss[loss=0.2605, simple_loss=0.3261, pruned_loss=0.09748, over 7196.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3314, pruned_loss=0.09834, over 1605088.66 frames. ], batch size: 16, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:17:48,716 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53259.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:17:52,100 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5898, 1.3509, 2.7603, 1.1529, 1.9006, 3.0154, 3.0092, 2.5684], + device='cuda:1'), covar=tensor([0.1027, 0.1368, 0.0440, 0.2019, 0.0799, 0.0301, 0.0517, 0.0624], + device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0276, 0.0231, 0.0272, 0.0243, 0.0218, 0.0276, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 04:17:56,642 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 04:17:59,374 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 04:18:09,733 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53288.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:18:16,610 INFO [train.py:901] (1/4) Epoch 7, batch 4800, loss[loss=0.2551, simple_loss=0.3037, pruned_loss=0.1033, over 7411.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3328, pruned_loss=0.0991, over 1608751.45 frames. ], batch size: 17, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:18:26,454 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53313.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:18:30,764 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-06 04:18:32,423 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:18:33,780 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3114, 1.6571, 4.4228, 1.8475, 2.3281, 5.2287, 5.0957, 4.5034], + device='cuda:1'), covar=tensor([0.1001, 0.1476, 0.0248, 0.1931, 0.0925, 0.0153, 0.0234, 0.0455], + device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0276, 0.0231, 0.0272, 0.0241, 0.0219, 0.0277, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 04:18:39,901 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53333.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:18:41,111 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.630e+02 3.191e+02 3.984e+02 9.617e+02, threshold=6.381e+02, percent-clipped=3.0 +2023-02-06 04:18:50,458 INFO [train.py:901] (1/4) Epoch 7, batch 4850, loss[loss=0.2361, simple_loss=0.3096, pruned_loss=0.08124, over 7965.00 frames. ], tot_loss[loss=0.2627, simple_loss=0.3305, pruned_loss=0.09743, over 1599447.06 frames. ], batch size: 21, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:18:51,164 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 04:19:12,830 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.08 vs. limit=5.0 +2023-02-06 04:19:16,472 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53385.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:19:26,613 INFO [train.py:901] (1/4) Epoch 7, batch 4900, loss[loss=0.2806, simple_loss=0.3595, pruned_loss=0.1008, over 8189.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.3308, pruned_loss=0.09756, over 1601516.58 frames. ], batch size: 23, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:19:38,409 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6468, 1.8025, 2.0324, 1.5980, 1.0444, 2.0863, 0.2272, 1.1637], + device='cuda:1'), covar=tensor([0.2943, 0.1521, 0.0624, 0.1948, 0.4844, 0.0563, 0.3674, 0.1904], + device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0142, 0.0084, 0.0190, 0.0228, 0.0088, 0.0147, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:19:40,217 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53418.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:19:40,949 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8017, 1.4619, 5.9026, 1.9786, 5.2322, 5.0253, 5.4952, 5.3253], + device='cuda:1'), covar=tensor([0.0414, 0.4093, 0.0203, 0.2738, 0.0780, 0.0540, 0.0365, 0.0393], + device='cuda:1'), in_proj_covar=tensor([0.0385, 0.0528, 0.0476, 0.0464, 0.0526, 0.0441, 0.0430, 0.0493], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 04:19:51,450 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 2.811e+02 3.269e+02 4.328e+02 9.769e+02, threshold=6.539e+02, percent-clipped=6.0 +2023-02-06 04:19:51,636 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7949, 1.5752, 5.8005, 2.0176, 5.2398, 4.9139, 5.4519, 5.2562], + device='cuda:1'), covar=tensor([0.0391, 0.4214, 0.0267, 0.2939, 0.0922, 0.0562, 0.0400, 0.0447], + device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0531, 0.0480, 0.0466, 0.0527, 0.0442, 0.0434, 0.0497], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 04:20:00,864 INFO [train.py:901] (1/4) Epoch 7, batch 4950, loss[loss=0.2912, simple_loss=0.357, pruned_loss=0.1127, over 8493.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3312, pruned_loss=0.09805, over 1604423.48 frames. ], batch size: 28, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:20:27,211 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53485.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:20:36,823 INFO [train.py:901] (1/4) Epoch 7, batch 5000, loss[loss=0.2522, simple_loss=0.3112, pruned_loss=0.09663, over 7433.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3309, pruned_loss=0.09739, over 1603772.21 frames. ], batch size: 17, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:20:45,268 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:21:01,965 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:21:03,200 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.612e+02 3.156e+02 4.000e+02 8.821e+02, threshold=6.312e+02, percent-clipped=7.0 +2023-02-06 04:21:12,890 INFO [train.py:901] (1/4) Epoch 7, batch 5050, loss[loss=0.247, simple_loss=0.3192, pruned_loss=0.08743, over 8439.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3299, pruned_loss=0.09635, over 1605440.26 frames. ], batch size: 29, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:21:32,025 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 04:21:46,781 INFO [train.py:901] (1/4) Epoch 7, batch 5100, loss[loss=0.294, simple_loss=0.3317, pruned_loss=0.1282, over 7784.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3299, pruned_loss=0.09707, over 1604401.96 frames. ], batch size: 19, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:21:51,220 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:21:56,055 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0694, 3.9672, 2.4646, 2.6613, 2.8525, 1.8681, 2.4753, 3.0091], + device='cuda:1'), covar=tensor([0.1524, 0.0269, 0.0887, 0.0723, 0.0650, 0.1174, 0.1060, 0.0884], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0241, 0.0316, 0.0304, 0.0315, 0.0322, 0.0348, 0.0324], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 04:22:13,233 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.889e+02 3.391e+02 4.238e+02 9.606e+02, threshold=6.783e+02, percent-clipped=10.0 +2023-02-06 04:22:23,450 INFO [train.py:901] (1/4) Epoch 7, batch 5150, loss[loss=0.3069, simple_loss=0.3594, pruned_loss=0.1272, over 7175.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3303, pruned_loss=0.09707, over 1606335.74 frames. ], batch size: 74, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:22:34,955 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53666.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:22:42,266 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53677.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:22:56,582 INFO [train.py:901] (1/4) Epoch 7, batch 5200, loss[loss=0.2402, simple_loss=0.3282, pruned_loss=0.07611, over 8459.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.3314, pruned_loss=0.09762, over 1614076.65 frames. ], batch size: 25, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:23:10,748 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:23:17,967 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53729.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:23:22,034 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.983e+02 3.078e+02 4.028e+02 5.378e+02 1.177e+03, threshold=8.056e+02, percent-clipped=8.0 +2023-02-06 04:23:28,940 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 04:23:32,186 INFO [train.py:901] (1/4) Epoch 7, batch 5250, loss[loss=0.2958, simple_loss=0.3527, pruned_loss=0.1195, over 8285.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3289, pruned_loss=0.09605, over 1611605.02 frames. ], batch size: 23, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:23:45,136 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0046, 2.2530, 3.7817, 1.6617, 2.9838, 2.4123, 2.0580, 2.6995], + device='cuda:1'), covar=tensor([0.1176, 0.1703, 0.0457, 0.2775, 0.0973, 0.1764, 0.1272, 0.1608], + device='cuda:1'), in_proj_covar=tensor([0.0480, 0.0477, 0.0529, 0.0554, 0.0603, 0.0531, 0.0451, 0.0591], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 04:23:54,706 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53781.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:24:00,274 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:24:02,338 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53792.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:24:06,925 INFO [train.py:901] (1/4) Epoch 7, batch 5300, loss[loss=0.3135, simple_loss=0.3745, pruned_loss=0.1262, over 8340.00 frames. ], tot_loss[loss=0.2617, simple_loss=0.3299, pruned_loss=0.09674, over 1613371.08 frames. ], batch size: 26, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:24:17,554 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53814.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:24:29,722 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 04:24:31,356 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.647e+02 3.169e+02 3.870e+02 1.211e+03, threshold=6.339e+02, percent-clipped=2.0 +2023-02-06 04:24:39,055 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53844.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:24:42,222 INFO [train.py:901] (1/4) Epoch 7, batch 5350, loss[loss=0.2373, simple_loss=0.3288, pruned_loss=0.07287, over 8461.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3301, pruned_loss=0.09658, over 1615827.59 frames. ], batch size: 27, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:25:17,625 INFO [train.py:901] (1/4) Epoch 7, batch 5400, loss[loss=0.2138, simple_loss=0.2865, pruned_loss=0.07054, over 7975.00 frames. ], tot_loss[loss=0.259, simple_loss=0.3272, pruned_loss=0.0954, over 1608376.84 frames. ], batch size: 21, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:25:28,186 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 +2023-02-06 04:25:41,926 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.728e+02 3.458e+02 4.119e+02 1.009e+03, threshold=6.915e+02, percent-clipped=3.0 +2023-02-06 04:25:51,251 INFO [train.py:901] (1/4) Epoch 7, batch 5450, loss[loss=0.2483, simple_loss=0.3195, pruned_loss=0.08858, over 8323.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3289, pruned_loss=0.09637, over 1610014.71 frames. ], batch size: 25, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:26:08,815 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53974.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:26:18,684 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 04:26:26,954 INFO [train.py:901] (1/4) Epoch 7, batch 5500, loss[loss=0.3139, simple_loss=0.3753, pruned_loss=0.1262, over 8806.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3293, pruned_loss=0.09659, over 1611542.83 frames. ], batch size: 32, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:26:27,124 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:26:35,489 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0122, 1.5522, 3.2498, 1.3191, 2.0203, 3.5429, 3.5600, 3.0787], + device='cuda:1'), covar=tensor([0.0941, 0.1472, 0.0362, 0.2035, 0.0931, 0.0272, 0.0404, 0.0668], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0272, 0.0233, 0.0269, 0.0237, 0.0219, 0.0274, 0.0279], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 04:26:52,074 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.794e+02 3.496e+02 4.646e+02 1.157e+03, threshold=6.993e+02, percent-clipped=7.0 +2023-02-06 04:26:53,580 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54037.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:01,157 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54048.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:01,618 INFO [train.py:901] (1/4) Epoch 7, batch 5550, loss[loss=0.2632, simple_loss=0.319, pruned_loss=0.1037, over 8245.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3289, pruned_loss=0.0965, over 1614175.00 frames. ], batch size: 22, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:27:10,519 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:11,404 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.92 vs. limit=5.0 +2023-02-06 04:27:18,030 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54073.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:27,406 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3861, 1.9598, 4.6509, 1.1315, 2.6310, 2.0914, 1.3839, 2.6368], + device='cuda:1'), covar=tensor([0.2156, 0.2553, 0.0673, 0.4391, 0.1735, 0.2954, 0.2270, 0.2501], + device='cuda:1'), in_proj_covar=tensor([0.0481, 0.0477, 0.0530, 0.0554, 0.0604, 0.0535, 0.0455, 0.0593], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-02-06 04:27:37,055 INFO [train.py:901] (1/4) Epoch 7, batch 5600, loss[loss=0.4122, simple_loss=0.4384, pruned_loss=0.193, over 6949.00 frames. ], tot_loss[loss=0.2602, simple_loss=0.3286, pruned_loss=0.09588, over 1613360.51 frames. ], batch size: 71, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:27:38,003 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54100.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:55,839 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:28:01,354 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3169, 1.9580, 3.0734, 2.5141, 2.6897, 1.9733, 1.5515, 1.3792], + device='cuda:1'), covar=tensor([0.2785, 0.3112, 0.0734, 0.1710, 0.1341, 0.1633, 0.1532, 0.3256], + device='cuda:1'), in_proj_covar=tensor([0.0828, 0.0779, 0.0670, 0.0769, 0.0858, 0.0709, 0.0657, 0.0707], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:28:02,460 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.954e+02 2.795e+02 3.455e+02 4.516e+02 9.788e+02, threshold=6.911e+02, percent-clipped=3.0 +2023-02-06 04:28:12,211 INFO [train.py:901] (1/4) Epoch 7, batch 5650, loss[loss=0.2453, simple_loss=0.3085, pruned_loss=0.09106, over 7426.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3292, pruned_loss=0.09607, over 1617648.86 frames. ], batch size: 17, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:28:25,370 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 04:28:25,791 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.70 vs. limit=5.0 +2023-02-06 04:28:26,283 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4007, 1.9856, 3.1200, 2.4227, 2.5977, 2.0048, 1.5493, 1.4041], + device='cuda:1'), covar=tensor([0.2357, 0.2748, 0.0701, 0.1602, 0.1322, 0.1404, 0.1296, 0.2823], + device='cuda:1'), in_proj_covar=tensor([0.0834, 0.0785, 0.0673, 0.0772, 0.0863, 0.0712, 0.0661, 0.0708], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:28:47,273 INFO [train.py:901] (1/4) Epoch 7, batch 5700, loss[loss=0.2719, simple_loss=0.3311, pruned_loss=0.1064, over 7920.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3287, pruned_loss=0.09593, over 1617757.82 frames. ], batch size: 20, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:28:54,428 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7749, 2.9740, 3.1893, 1.9050, 1.4780, 3.5397, 0.6795, 2.1227], + device='cuda:1'), covar=tensor([0.3252, 0.1423, 0.0465, 0.3299, 0.6259, 0.0301, 0.5165, 0.1998], + device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0143, 0.0086, 0.0192, 0.0228, 0.0088, 0.0151, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:29:12,993 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.786e+02 3.155e+02 4.023e+02 8.991e+02, threshold=6.311e+02, percent-clipped=4.0 +2023-02-06 04:29:22,469 INFO [train.py:901] (1/4) Epoch 7, batch 5750, loss[loss=0.2543, simple_loss=0.3268, pruned_loss=0.09087, over 8512.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3295, pruned_loss=0.09658, over 1613688.37 frames. ], batch size: 26, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:29:31,472 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 04:29:40,474 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2464, 1.8443, 2.9484, 2.3008, 2.5332, 1.9700, 1.5815, 1.2032], + device='cuda:1'), covar=tensor([0.2587, 0.2792, 0.0645, 0.1601, 0.1254, 0.1506, 0.1337, 0.3060], + device='cuda:1'), in_proj_covar=tensor([0.0822, 0.0771, 0.0660, 0.0760, 0.0852, 0.0703, 0.0653, 0.0700], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:29:54,386 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0375, 1.6728, 1.6827, 1.5664, 1.3322, 1.6535, 2.2685, 1.9376], + device='cuda:1'), covar=tensor([0.0450, 0.1226, 0.1817, 0.1391, 0.0560, 0.1465, 0.0621, 0.0566], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0167, 0.0209, 0.0171, 0.0116, 0.0174, 0.0128, 0.0146], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 04:29:56,269 INFO [train.py:901] (1/4) Epoch 7, batch 5800, loss[loss=0.2362, simple_loss=0.3318, pruned_loss=0.07028, over 8290.00 frames. ], tot_loss[loss=0.2614, simple_loss=0.3294, pruned_loss=0.09669, over 1611814.82 frames. ], batch size: 23, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:30:18,113 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7850, 1.9677, 1.6114, 2.3347, 1.0633, 1.3891, 1.6471, 1.9593], + device='cuda:1'), covar=tensor([0.0798, 0.0856, 0.1207, 0.0583, 0.1254, 0.1698, 0.1042, 0.0865], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0237, 0.0272, 0.0224, 0.0235, 0.0265, 0.0275, 0.0238], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 04:30:22,037 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.127e+02 2.882e+02 3.737e+02 4.385e+02 9.194e+02, threshold=7.474e+02, percent-clipped=5.0 +2023-02-06 04:30:32,284 INFO [train.py:901] (1/4) Epoch 7, batch 5850, loss[loss=0.2575, simple_loss=0.3306, pruned_loss=0.0922, over 8135.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3296, pruned_loss=0.09681, over 1611279.00 frames. ], batch size: 22, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:30:40,665 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9686, 3.1126, 3.2353, 2.1513, 1.6843, 3.4993, 0.6979, 2.1756], + device='cuda:1'), covar=tensor([0.2677, 0.1150, 0.0469, 0.2994, 0.5149, 0.0374, 0.4933, 0.1870], + device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0144, 0.0086, 0.0193, 0.0230, 0.0088, 0.0152, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:31:06,118 INFO [train.py:901] (1/4) Epoch 7, batch 5900, loss[loss=0.2772, simple_loss=0.3476, pruned_loss=0.1034, over 8106.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3292, pruned_loss=0.09662, over 1611991.66 frames. ], batch size: 23, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:31:30,659 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.634e+02 3.151e+02 3.851e+02 7.879e+02, threshold=6.301e+02, percent-clipped=2.0 +2023-02-06 04:31:40,696 INFO [train.py:901] (1/4) Epoch 7, batch 5950, loss[loss=0.1853, simple_loss=0.2568, pruned_loss=0.05693, over 7711.00 frames. ], tot_loss[loss=0.2615, simple_loss=0.3292, pruned_loss=0.09689, over 1605649.49 frames. ], batch size: 18, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:32:13,860 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4946, 2.0430, 2.0746, 1.1977, 2.1712, 1.5785, 0.5789, 1.8545], + device='cuda:1'), covar=tensor([0.0266, 0.0129, 0.0098, 0.0204, 0.0152, 0.0385, 0.0349, 0.0108], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0262, 0.0223, 0.0325, 0.0263, 0.0414, 0.0322, 0.0300], + device='cuda:1'), out_proj_covar=tensor([1.1091e-04, 8.1536e-05, 6.8065e-05, 1.0005e-04, 8.2443e-05, 1.3947e-04, + 1.0222e-04, 9.3704e-05], device='cuda:1') +2023-02-06 04:32:14,309 INFO [train.py:901] (1/4) Epoch 7, batch 6000, loss[loss=0.233, simple_loss=0.2986, pruned_loss=0.08373, over 7932.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3293, pruned_loss=0.09666, over 1607196.25 frames. ], batch size: 20, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:32:14,309 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 04:32:26,542 INFO [train.py:935] (1/4) Epoch 7, validation: loss=0.2048, simple_loss=0.3036, pruned_loss=0.05298, over 944034.00 frames. +2023-02-06 04:32:26,543 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6607MB +2023-02-06 04:32:41,083 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3264, 1.6617, 1.6762, 1.3545, 0.9667, 1.4778, 1.7511, 1.8797], + device='cuda:1'), covar=tensor([0.0530, 0.1252, 0.1763, 0.1387, 0.0630, 0.1508, 0.0712, 0.0573], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0165, 0.0206, 0.0168, 0.0115, 0.0171, 0.0127, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 04:32:50,868 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.687e+02 3.524e+02 4.445e+02 8.914e+02, threshold=7.048e+02, percent-clipped=8.0 +2023-02-06 04:33:00,127 INFO [train.py:901] (1/4) Epoch 7, batch 6050, loss[loss=0.2668, simple_loss=0.3219, pruned_loss=0.1058, over 7923.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3288, pruned_loss=0.0965, over 1607395.07 frames. ], batch size: 20, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:33:04,706 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 04:33:17,078 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4050, 2.2624, 1.6488, 2.0203, 1.9473, 1.3544, 1.7172, 1.7918], + device='cuda:1'), covar=tensor([0.0957, 0.0280, 0.0873, 0.0400, 0.0569, 0.1163, 0.0686, 0.0618], + device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0234, 0.0309, 0.0299, 0.0313, 0.0314, 0.0334, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 04:33:36,271 INFO [train.py:901] (1/4) Epoch 7, batch 6100, loss[loss=0.2432, simple_loss=0.3116, pruned_loss=0.08739, over 7825.00 frames. ], tot_loss[loss=0.2615, simple_loss=0.3292, pruned_loss=0.09684, over 1611169.41 frames. ], batch size: 20, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:33:44,373 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 04:34:00,453 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 04:34:01,818 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.824e+02 3.447e+02 4.351e+02 1.012e+03, threshold=6.894e+02, percent-clipped=2.0 +2023-02-06 04:34:08,855 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3327, 1.5775, 2.2454, 1.1997, 1.5903, 1.6414, 1.3705, 1.5220], + device='cuda:1'), covar=tensor([0.1480, 0.1607, 0.0636, 0.3001, 0.1261, 0.2263, 0.1576, 0.1561], + device='cuda:1'), in_proj_covar=tensor([0.0478, 0.0473, 0.0526, 0.0558, 0.0599, 0.0530, 0.0453, 0.0588], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:34:11,167 INFO [train.py:901] (1/4) Epoch 7, batch 6150, loss[loss=0.21, simple_loss=0.2844, pruned_loss=0.06777, over 7411.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3298, pruned_loss=0.09715, over 1614938.32 frames. ], batch size: 17, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:34:34,031 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=54682.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:34:46,669 INFO [train.py:901] (1/4) Epoch 7, batch 6200, loss[loss=0.2725, simple_loss=0.3444, pruned_loss=0.1003, over 8434.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3306, pruned_loss=0.09749, over 1616484.26 frames. ], batch size: 27, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:34:49,810 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.59 vs. limit=5.0 +2023-02-06 04:35:12,147 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.928e+02 3.624e+02 4.953e+02 9.267e+02, threshold=7.248e+02, percent-clipped=4.0 +2023-02-06 04:35:21,786 INFO [train.py:901] (1/4) Epoch 7, batch 6250, loss[loss=0.2571, simple_loss=0.326, pruned_loss=0.09408, over 7974.00 frames. ], tot_loss[loss=0.2626, simple_loss=0.3308, pruned_loss=0.09721, over 1617730.19 frames. ], batch size: 21, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:35:40,874 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1149, 2.3773, 1.8010, 2.7539, 1.3861, 1.6009, 1.7195, 2.2706], + device='cuda:1'), covar=tensor([0.0694, 0.0707, 0.1216, 0.0487, 0.1219, 0.1514, 0.1313, 0.0767], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0235, 0.0274, 0.0225, 0.0235, 0.0264, 0.0273, 0.0238], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 04:35:55,506 INFO [train.py:901] (1/4) Epoch 7, batch 6300, loss[loss=0.3319, simple_loss=0.3872, pruned_loss=0.1382, over 8340.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3298, pruned_loss=0.09717, over 1608662.33 frames. ], batch size: 25, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:36:22,283 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.742e+02 3.001e+02 3.662e+02 4.451e+02 9.002e+02, threshold=7.325e+02, percent-clipped=3.0 +2023-02-06 04:36:32,313 INFO [train.py:901] (1/4) Epoch 7, batch 6350, loss[loss=0.2453, simple_loss=0.3348, pruned_loss=0.07791, over 8566.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3294, pruned_loss=0.09743, over 1605701.60 frames. ], batch size: 31, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:36:53,641 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=54880.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:37:06,004 INFO [train.py:901] (1/4) Epoch 7, batch 6400, loss[loss=0.2744, simple_loss=0.3504, pruned_loss=0.09918, over 8658.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3292, pruned_loss=0.09655, over 1609269.29 frames. ], batch size: 34, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:37:15,525 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.52 vs. limit=5.0 +2023-02-06 04:37:31,198 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.661e+02 3.281e+02 3.949e+02 1.010e+03, threshold=6.562e+02, percent-clipped=2.0 +2023-02-06 04:37:40,720 INFO [train.py:901] (1/4) Epoch 7, batch 6450, loss[loss=0.2088, simple_loss=0.2862, pruned_loss=0.06564, over 7925.00 frames. ], tot_loss[loss=0.2614, simple_loss=0.3295, pruned_loss=0.09664, over 1609311.41 frames. ], batch size: 20, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:37:47,759 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0575, 1.2807, 1.1462, 0.4369, 1.1593, 0.9017, 0.2068, 0.9993], + device='cuda:1'), covar=tensor([0.0228, 0.0162, 0.0154, 0.0286, 0.0204, 0.0494, 0.0359, 0.0163], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0261, 0.0221, 0.0322, 0.0261, 0.0411, 0.0320, 0.0299], + device='cuda:1'), out_proj_covar=tensor([1.1000e-04, 8.0842e-05, 6.7190e-05, 9.8893e-05, 8.1659e-05, 1.3821e-04, + 1.0117e-04, 9.3070e-05], device='cuda:1') +2023-02-06 04:37:50,649 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-06 04:37:57,799 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0124, 1.8913, 1.7337, 1.4956, 1.1464, 1.5185, 2.3125, 2.1777], + device='cuda:1'), covar=tensor([0.0470, 0.1119, 0.1705, 0.1371, 0.0589, 0.1443, 0.0632, 0.0527], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0165, 0.0206, 0.0168, 0.0115, 0.0171, 0.0126, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 04:38:15,593 INFO [train.py:901] (1/4) Epoch 7, batch 6500, loss[loss=0.2382, simple_loss=0.3224, pruned_loss=0.07693, over 8369.00 frames. ], tot_loss[loss=0.2595, simple_loss=0.3284, pruned_loss=0.09528, over 1610535.83 frames. ], batch size: 24, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:38:33,806 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55026.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:38:39,665 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.546e+02 3.271e+02 4.197e+02 5.859e+02, threshold=6.542e+02, percent-clipped=0.0 +2023-02-06 04:38:49,577 INFO [train.py:901] (1/4) Epoch 7, batch 6550, loss[loss=0.2615, simple_loss=0.3342, pruned_loss=0.09438, over 8677.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3293, pruned_loss=0.09618, over 1610632.62 frames. ], batch size: 39, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:39:04,762 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9841, 1.8368, 3.3337, 1.3522, 2.1753, 3.6940, 3.5894, 3.1260], + device='cuda:1'), covar=tensor([0.0981, 0.1359, 0.0336, 0.2141, 0.0945, 0.0272, 0.0445, 0.0594], + device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0278, 0.0233, 0.0272, 0.0245, 0.0219, 0.0279, 0.0283], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 04:39:12,186 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 04:39:25,807 INFO [train.py:901] (1/4) Epoch 7, batch 6600, loss[loss=0.2544, simple_loss=0.322, pruned_loss=0.09337, over 8520.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3296, pruned_loss=0.09591, over 1611122.20 frames. ], batch size: 26, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:39:32,339 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 04:39:49,418 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.699e+02 3.503e+02 4.413e+02 7.218e+02, threshold=7.007e+02, percent-clipped=4.0 +2023-02-06 04:39:53,564 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55141.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:39:58,774 INFO [train.py:901] (1/4) Epoch 7, batch 6650, loss[loss=0.2609, simple_loss=0.3426, pruned_loss=0.08966, over 8489.00 frames. ], tot_loss[loss=0.2598, simple_loss=0.3286, pruned_loss=0.09552, over 1606715.62 frames. ], batch size: 39, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:40:10,779 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55166.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:40:16,890 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2843, 1.4489, 4.3235, 1.9002, 2.4746, 4.9709, 4.9725, 4.3595], + device='cuda:1'), covar=tensor([0.1031, 0.1618, 0.0272, 0.1959, 0.0893, 0.0214, 0.0360, 0.0593], + device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0275, 0.0233, 0.0272, 0.0242, 0.0216, 0.0275, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 04:40:34,123 INFO [train.py:901] (1/4) Epoch 7, batch 6700, loss[loss=0.2137, simple_loss=0.2855, pruned_loss=0.071, over 7935.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3298, pruned_loss=0.09594, over 1611488.20 frames. ], batch size: 20, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:40:42,738 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-06 04:40:51,581 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55224.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:40:55,052 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55229.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:40:58,796 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 3.031e+02 3.759e+02 4.673e+02 1.170e+03, threshold=7.519e+02, percent-clipped=9.0 +2023-02-06 04:41:07,986 INFO [train.py:901] (1/4) Epoch 7, batch 6750, loss[loss=0.2878, simple_loss=0.3419, pruned_loss=0.1168, over 7536.00 frames. ], tot_loss[loss=0.2615, simple_loss=0.3301, pruned_loss=0.09649, over 1608851.90 frames. ], batch size: 18, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:41:42,535 INFO [train.py:901] (1/4) Epoch 7, batch 6800, loss[loss=0.2865, simple_loss=0.3685, pruned_loss=0.1022, over 8473.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3306, pruned_loss=0.0968, over 1607039.22 frames. ], batch size: 25, lr: 1.07e-02, grad_scale: 16.0 +2023-02-06 04:41:47,245 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 04:41:57,717 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1796, 3.0710, 2.8572, 1.6085, 2.7534, 2.8144, 2.8713, 2.6427], + device='cuda:1'), covar=tensor([0.1181, 0.0911, 0.1320, 0.4416, 0.1059, 0.1162, 0.1446, 0.1331], + device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0323, 0.0356, 0.0434, 0.0340, 0.0320, 0.0328, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:42:08,557 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.859e+02 3.364e+02 4.161e+02 9.626e+02, threshold=6.728e+02, percent-clipped=3.0 +2023-02-06 04:42:11,503 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55339.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:42:18,201 INFO [train.py:901] (1/4) Epoch 7, batch 6850, loss[loss=0.2711, simple_loss=0.3368, pruned_loss=0.1027, over 8280.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3297, pruned_loss=0.09635, over 1610210.69 frames. ], batch size: 23, lr: 1.06e-02, grad_scale: 16.0 +2023-02-06 04:42:24,371 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7386, 3.1834, 2.2815, 4.0775, 1.9282, 2.2172, 2.3407, 2.9137], + device='cuda:1'), covar=tensor([0.0841, 0.0867, 0.1275, 0.0294, 0.1244, 0.1605, 0.1423, 0.1149], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0233, 0.0272, 0.0221, 0.0230, 0.0265, 0.0274, 0.0236], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 04:42:34,179 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 04:42:50,630 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55397.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:42:52,432 INFO [train.py:901] (1/4) Epoch 7, batch 6900, loss[loss=0.3099, simple_loss=0.3564, pruned_loss=0.1317, over 7970.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.3293, pruned_loss=0.09641, over 1609512.94 frames. ], batch size: 21, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:43:04,239 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 +2023-02-06 04:43:06,187 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1999, 1.8161, 1.5419, 1.4511, 1.3304, 1.6713, 2.1726, 1.8057], + device='cuda:1'), covar=tensor([0.0446, 0.1349, 0.1854, 0.1498, 0.0605, 0.1551, 0.0690, 0.0608], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0166, 0.0204, 0.0168, 0.0114, 0.0170, 0.0125, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 04:43:09,642 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55422.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:43:19,279 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.767e+02 3.318e+02 4.413e+02 7.718e+02, threshold=6.635e+02, percent-clipped=1.0 +2023-02-06 04:43:19,733 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 04:43:28,924 INFO [train.py:901] (1/4) Epoch 7, batch 6950, loss[loss=0.2828, simple_loss=0.3526, pruned_loss=0.1065, over 8137.00 frames. ], tot_loss[loss=0.2601, simple_loss=0.3287, pruned_loss=0.09579, over 1608373.42 frames. ], batch size: 22, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:43:42,918 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 04:43:46,600 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 04:43:48,202 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-06 04:44:02,247 INFO [train.py:901] (1/4) Epoch 7, batch 7000, loss[loss=0.2242, simple_loss=0.2898, pruned_loss=0.07933, over 7231.00 frames. ], tot_loss[loss=0.2593, simple_loss=0.3275, pruned_loss=0.09557, over 1602150.99 frames. ], batch size: 16, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:44:04,367 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5279, 2.7683, 1.5858, 2.1979, 2.0993, 1.3254, 1.9645, 2.0825], + device='cuda:1'), covar=tensor([0.1222, 0.0240, 0.0983, 0.0540, 0.0619, 0.1262, 0.0855, 0.0821], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0232, 0.0308, 0.0297, 0.0308, 0.0317, 0.0335, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 04:44:09,453 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:44:28,238 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.942e+02 3.699e+02 4.542e+02 1.220e+03, threshold=7.399e+02, percent-clipped=11.0 +2023-02-06 04:44:37,061 INFO [train.py:901] (1/4) Epoch 7, batch 7050, loss[loss=0.2388, simple_loss=0.3259, pruned_loss=0.07586, over 8506.00 frames. ], tot_loss[loss=0.2594, simple_loss=0.3279, pruned_loss=0.09545, over 1607335.08 frames. ], batch size: 26, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:44:43,035 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 04:44:54,228 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:45:09,303 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55595.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:45:11,744 INFO [train.py:901] (1/4) Epoch 7, batch 7100, loss[loss=0.229, simple_loss=0.302, pruned_loss=0.07798, over 7977.00 frames. ], tot_loss[loss=0.259, simple_loss=0.3278, pruned_loss=0.09509, over 1608055.08 frames. ], batch size: 21, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:45:26,282 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:45:29,489 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55625.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:45:30,152 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7019, 1.4650, 3.0519, 1.1556, 2.2138, 3.2896, 3.2961, 2.8006], + device='cuda:1'), covar=tensor([0.1032, 0.1322, 0.0367, 0.1920, 0.0785, 0.0290, 0.0392, 0.0606], + device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0270, 0.0229, 0.0268, 0.0239, 0.0212, 0.0277, 0.0277], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 04:45:36,697 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.720e+02 3.297e+02 4.008e+02 7.250e+02, threshold=6.594e+02, percent-clipped=0.0 +2023-02-06 04:45:45,958 INFO [train.py:901] (1/4) Epoch 7, batch 7150, loss[loss=0.2513, simple_loss=0.3151, pruned_loss=0.09381, over 7250.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3272, pruned_loss=0.09472, over 1605008.45 frames. ], batch size: 16, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:45:57,609 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6081, 1.2756, 4.7671, 1.9263, 4.0945, 3.9939, 4.3579, 4.2295], + device='cuda:1'), covar=tensor([0.0424, 0.3905, 0.0348, 0.2652, 0.1085, 0.0654, 0.0397, 0.0483], + device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0531, 0.0481, 0.0464, 0.0540, 0.0442, 0.0445, 0.0503], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 04:46:14,244 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55688.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:46:21,671 INFO [train.py:901] (1/4) Epoch 7, batch 7200, loss[loss=0.2676, simple_loss=0.3379, pruned_loss=0.0987, over 8730.00 frames. ], tot_loss[loss=0.2568, simple_loss=0.3258, pruned_loss=0.09392, over 1604114.56 frames. ], batch size: 49, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:46:21,923 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5679, 2.1638, 3.3957, 1.2940, 2.3240, 1.9953, 1.6397, 2.2176], + device='cuda:1'), covar=tensor([0.1545, 0.1822, 0.0500, 0.3447, 0.1352, 0.2411, 0.1574, 0.1999], + device='cuda:1'), in_proj_covar=tensor([0.0476, 0.0478, 0.0534, 0.0561, 0.0599, 0.0535, 0.0455, 0.0594], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:46:33,577 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6969, 1.8544, 2.1458, 1.7283, 1.0191, 2.2528, 0.3150, 1.2557], + device='cuda:1'), covar=tensor([0.2755, 0.1662, 0.0558, 0.1817, 0.5011, 0.0643, 0.4027, 0.2025], + device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0145, 0.0085, 0.0192, 0.0232, 0.0090, 0.0146, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:46:38,473 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9181, 4.1290, 2.3320, 2.6304, 2.8531, 2.1285, 2.6919, 2.9839], + device='cuda:1'), covar=tensor([0.1514, 0.0187, 0.0941, 0.0704, 0.0666, 0.1171, 0.0974, 0.0937], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0232, 0.0307, 0.0298, 0.0309, 0.0316, 0.0337, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 04:46:47,138 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.857e+02 3.487e+02 4.455e+02 1.230e+03, threshold=6.974e+02, percent-clipped=5.0 +2023-02-06 04:46:55,812 INFO [train.py:901] (1/4) Epoch 7, batch 7250, loss[loss=0.2973, simple_loss=0.3569, pruned_loss=0.1189, over 8137.00 frames. ], tot_loss[loss=0.2592, simple_loss=0.328, pruned_loss=0.09519, over 1607047.45 frames. ], batch size: 22, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:47:07,787 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55766.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:47:31,018 INFO [train.py:901] (1/4) Epoch 7, batch 7300, loss[loss=0.2258, simple_loss=0.3011, pruned_loss=0.07523, over 8134.00 frames. ], tot_loss[loss=0.2602, simple_loss=0.3285, pruned_loss=0.09592, over 1604715.57 frames. ], batch size: 22, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:47:55,727 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.887e+02 3.402e+02 4.424e+02 1.529e+03, threshold=6.804e+02, percent-clipped=7.0 +2023-02-06 04:48:04,224 INFO [train.py:901] (1/4) Epoch 7, batch 7350, loss[loss=0.2411, simple_loss=0.3132, pruned_loss=0.08452, over 8238.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.329, pruned_loss=0.0963, over 1608353.10 frames. ], batch size: 22, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:48:15,834 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55866.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:48:26,624 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55881.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:48:27,824 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 04:48:40,240 INFO [train.py:901] (1/4) Epoch 7, batch 7400, loss[loss=0.2651, simple_loss=0.3496, pruned_loss=0.09031, over 8103.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3301, pruned_loss=0.09704, over 1608088.71 frames. ], batch size: 23, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:48:45,329 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55906.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:48:50,034 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 04:49:01,383 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9042, 1.5123, 1.4944, 1.3177, 1.1289, 1.3075, 1.5659, 1.6235], + device='cuda:1'), covar=tensor([0.0535, 0.1257, 0.1636, 0.1352, 0.0552, 0.1520, 0.0712, 0.0578], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0166, 0.0204, 0.0168, 0.0113, 0.0172, 0.0127, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 04:49:05,902 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.646e+02 3.471e+02 4.467e+02 1.348e+03, threshold=6.942e+02, percent-clipped=5.0 +2023-02-06 04:49:11,671 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:49:14,923 INFO [train.py:901] (1/4) Epoch 7, batch 7450, loss[loss=0.2529, simple_loss=0.3232, pruned_loss=0.09128, over 7816.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3305, pruned_loss=0.09695, over 1609511.79 frames. ], batch size: 20, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:49:25,422 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6538, 2.8475, 1.7913, 2.2720, 2.4472, 1.5609, 2.0525, 2.1903], + device='cuda:1'), covar=tensor([0.1174, 0.0266, 0.0915, 0.0566, 0.0577, 0.1181, 0.0888, 0.0806], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0233, 0.0309, 0.0301, 0.0310, 0.0319, 0.0334, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 04:49:25,911 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 04:49:28,832 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:49:38,426 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6730, 1.7741, 2.0841, 1.7149, 1.2212, 2.2199, 0.2109, 1.2484], + device='cuda:1'), covar=tensor([0.3160, 0.1715, 0.0670, 0.2025, 0.4799, 0.0615, 0.4545, 0.2253], + device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0144, 0.0085, 0.0191, 0.0229, 0.0088, 0.0145, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:49:50,667 INFO [train.py:901] (1/4) Epoch 7, batch 7500, loss[loss=0.2292, simple_loss=0.2972, pruned_loss=0.08067, over 8025.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3315, pruned_loss=0.09745, over 1613764.78 frames. ], batch size: 22, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:50:17,129 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.816e+02 3.537e+02 4.737e+02 9.745e+02, threshold=7.074e+02, percent-clipped=6.0 +2023-02-06 04:50:23,239 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5218, 4.5027, 4.0261, 1.9210, 3.9292, 4.0133, 4.0749, 3.5202], + device='cuda:1'), covar=tensor([0.0769, 0.0616, 0.1044, 0.4971, 0.0803, 0.0798, 0.1318, 0.0992], + device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0330, 0.0361, 0.0447, 0.0345, 0.0320, 0.0335, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:50:25,625 INFO [train.py:901] (1/4) Epoch 7, batch 7550, loss[loss=0.2978, simple_loss=0.3573, pruned_loss=0.1191, over 8284.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.331, pruned_loss=0.09702, over 1614308.93 frames. ], batch size: 23, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:50:30,389 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1069, 2.5169, 1.8914, 2.9989, 1.6213, 1.6699, 2.3265, 2.4592], + device='cuda:1'), covar=tensor([0.0846, 0.0933, 0.1278, 0.0377, 0.1220, 0.1679, 0.1005, 0.0880], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0238, 0.0278, 0.0222, 0.0233, 0.0270, 0.0274, 0.0238], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 04:50:58,638 INFO [train.py:901] (1/4) Epoch 7, batch 7600, loss[loss=0.2907, simple_loss=0.3496, pruned_loss=0.1159, over 8344.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3314, pruned_loss=0.09788, over 1615577.58 frames. ], batch size: 26, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:51:06,779 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56110.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:51:25,295 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 2.750e+02 3.495e+02 4.537e+02 9.121e+02, threshold=6.990e+02, percent-clipped=3.0 +2023-02-06 04:51:34,926 INFO [train.py:901] (1/4) Epoch 7, batch 7650, loss[loss=0.2585, simple_loss=0.3295, pruned_loss=0.09375, over 8466.00 frames. ], tot_loss[loss=0.2642, simple_loss=0.3317, pruned_loss=0.09832, over 1617941.80 frames. ], batch size: 29, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:51:44,425 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56163.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:51:47,916 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4820, 2.0070, 3.1264, 2.3867, 2.6105, 2.0154, 1.4858, 1.3933], + device='cuda:1'), covar=tensor([0.2644, 0.3195, 0.0732, 0.1912, 0.1489, 0.1633, 0.1659, 0.3092], + device='cuda:1'), in_proj_covar=tensor([0.0838, 0.0784, 0.0674, 0.0785, 0.0869, 0.0725, 0.0677, 0.0713], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:52:08,628 INFO [train.py:901] (1/4) Epoch 7, batch 7700, loss[loss=0.3351, simple_loss=0.385, pruned_loss=0.1426, over 7089.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3314, pruned_loss=0.09794, over 1617665.14 frames. ], batch size: 71, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:52:16,094 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56210.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:52:26,762 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56225.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:52:34,688 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.791e+02 3.394e+02 3.978e+02 9.035e+02, threshold=6.788e+02, percent-clipped=3.0 +2023-02-06 04:52:34,717 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 04:52:44,019 INFO [train.py:901] (1/4) Epoch 7, batch 7750, loss[loss=0.2966, simple_loss=0.3679, pruned_loss=0.1127, over 8325.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3311, pruned_loss=0.09747, over 1619376.76 frames. ], batch size: 25, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:52:54,605 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.31 vs. limit=5.0 +2023-02-06 04:53:18,264 INFO [train.py:901] (1/4) Epoch 7, batch 7800, loss[loss=0.2271, simple_loss=0.3025, pruned_loss=0.07585, over 7813.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3319, pruned_loss=0.09765, over 1622888.63 frames. ], batch size: 20, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:53:22,037 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 04:53:35,675 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56325.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:53:36,868 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8727, 3.7985, 3.4620, 1.4895, 3.3483, 3.1630, 3.4760, 2.8725], + device='cuda:1'), covar=tensor([0.0967, 0.0797, 0.1132, 0.5271, 0.0974, 0.1172, 0.1703, 0.1162], + device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0331, 0.0364, 0.0455, 0.0351, 0.0324, 0.0336, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:53:42,612 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.702e+02 3.307e+02 4.383e+02 8.490e+02, threshold=6.613e+02, percent-clipped=4.0 +2023-02-06 04:53:51,371 INFO [train.py:901] (1/4) Epoch 7, batch 7850, loss[loss=0.2819, simple_loss=0.3588, pruned_loss=0.1025, over 8510.00 frames. ], tot_loss[loss=0.2649, simple_loss=0.3331, pruned_loss=0.09838, over 1620893.85 frames. ], batch size: 26, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:53:54,171 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1246, 3.0052, 2.8447, 1.5441, 2.7348, 2.7932, 2.9067, 2.5630], + device='cuda:1'), covar=tensor([0.1438, 0.0993, 0.1428, 0.5266, 0.1208, 0.1407, 0.1668, 0.1363], + device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0331, 0.0364, 0.0454, 0.0349, 0.0326, 0.0337, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:54:09,406 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 04:54:24,869 INFO [train.py:901] (1/4) Epoch 7, batch 7900, loss[loss=0.3059, simple_loss=0.3826, pruned_loss=0.1146, over 8331.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3355, pruned_loss=0.1, over 1616310.70 frames. ], batch size: 25, lr: 1.05e-02, grad_scale: 8.0 +2023-02-06 04:54:27,218 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3521, 2.0412, 3.2035, 2.5101, 2.7914, 2.0529, 1.5532, 1.3466], + device='cuda:1'), covar=tensor([0.2860, 0.3062, 0.0722, 0.1762, 0.1429, 0.1679, 0.1394, 0.3216], + device='cuda:1'), in_proj_covar=tensor([0.0841, 0.0786, 0.0677, 0.0786, 0.0876, 0.0732, 0.0677, 0.0713], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:54:49,425 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.758e+02 3.520e+02 4.424e+02 1.197e+03, threshold=7.039e+02, percent-clipped=9.0 +2023-02-06 04:54:58,048 INFO [train.py:901] (1/4) Epoch 7, batch 7950, loss[loss=0.2414, simple_loss=0.3131, pruned_loss=0.08483, over 8229.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3345, pruned_loss=0.09926, over 1612099.55 frames. ], batch size: 22, lr: 1.05e-02, grad_scale: 8.0 +2023-02-06 04:55:19,800 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56481.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:55:31,645 INFO [train.py:901] (1/4) Epoch 7, batch 8000, loss[loss=0.2624, simple_loss=0.3412, pruned_loss=0.0918, over 8108.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3336, pruned_loss=0.09847, over 1613807.64 frames. ], batch size: 23, lr: 1.05e-02, grad_scale: 8.0 +2023-02-06 04:55:36,306 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56506.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:55:36,850 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:55:56,108 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.809e+02 3.378e+02 4.457e+02 7.052e+02, threshold=6.755e+02, percent-clipped=1.0 +2023-02-06 04:55:59,635 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56541.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:56:04,837 INFO [train.py:901] (1/4) Epoch 7, batch 8050, loss[loss=0.3269, simple_loss=0.3734, pruned_loss=0.1402, over 6891.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3324, pruned_loss=0.09877, over 1602126.78 frames. ], batch size: 71, lr: 1.05e-02, grad_scale: 8.0 +2023-02-06 04:56:37,708 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 04:56:42,633 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56581.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:56:43,099 INFO [train.py:901] (1/4) Epoch 8, batch 0, loss[loss=0.2558, simple_loss=0.327, pruned_loss=0.09228, over 8558.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.327, pruned_loss=0.09228, over 8558.00 frames. ], batch size: 31, lr: 9.92e-03, grad_scale: 8.0 +2023-02-06 04:56:43,099 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 04:56:54,074 INFO [train.py:935] (1/4) Epoch 8, validation: loss=0.205, simple_loss=0.3028, pruned_loss=0.05355, over 944034.00 frames. +2023-02-06 04:56:54,074 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6607MB +2023-02-06 04:56:54,916 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2884, 4.2647, 3.8675, 1.4982, 3.7870, 3.8979, 3.9411, 3.6281], + device='cuda:1'), covar=tensor([0.1045, 0.0698, 0.1212, 0.6179, 0.0843, 0.1099, 0.1356, 0.0951], + device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0327, 0.0354, 0.0451, 0.0343, 0.0322, 0.0334, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:57:08,614 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 04:57:10,771 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56606.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:57:22,349 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:57:28,998 INFO [train.py:901] (1/4) Epoch 8, batch 50, loss[loss=0.3288, simple_loss=0.3639, pruned_loss=0.1468, over 6981.00 frames. ], tot_loss[loss=0.2675, simple_loss=0.3358, pruned_loss=0.09957, over 368996.56 frames. ], batch size: 72, lr: 9.92e-03, grad_scale: 8.0 +2023-02-06 04:57:31,767 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.799e+02 2.831e+02 3.488e+02 4.265e+02 1.069e+03, threshold=6.975e+02, percent-clipped=2.0 +2023-02-06 04:57:39,013 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-02-06 04:57:43,161 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 04:58:03,657 INFO [train.py:901] (1/4) Epoch 8, batch 100, loss[loss=0.2787, simple_loss=0.3399, pruned_loss=0.1087, over 8640.00 frames. ], tot_loss[loss=0.2666, simple_loss=0.3342, pruned_loss=0.09946, over 646599.27 frames. ], batch size: 31, lr: 9.91e-03, grad_scale: 8.0 +2023-02-06 04:58:05,727 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 04:58:38,295 INFO [train.py:901] (1/4) Epoch 8, batch 150, loss[loss=0.2768, simple_loss=0.3585, pruned_loss=0.09753, over 8323.00 frames. ], tot_loss[loss=0.2635, simple_loss=0.3318, pruned_loss=0.09755, over 860551.26 frames. ], batch size: 25, lr: 9.91e-03, grad_scale: 8.0 +2023-02-06 04:58:40,473 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([6.0922, 1.6577, 6.1218, 1.9657, 5.4751, 5.2252, 5.7090, 5.4935], + device='cuda:1'), covar=tensor([0.0322, 0.3591, 0.0238, 0.2988, 0.0857, 0.0681, 0.0301, 0.0420], + device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0531, 0.0489, 0.0473, 0.0538, 0.0450, 0.0448, 0.0507], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 04:58:40,997 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.710e+02 3.372e+02 4.105e+02 8.611e+02, threshold=6.744e+02, percent-clipped=2.0 +2023-02-06 04:59:12,801 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 04:59:13,819 INFO [train.py:901] (1/4) Epoch 8, batch 200, loss[loss=0.2575, simple_loss=0.3304, pruned_loss=0.09226, over 8077.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3322, pruned_loss=0.09712, over 1032506.37 frames. ], batch size: 21, lr: 9.90e-03, grad_scale: 8.0 +2023-02-06 04:59:41,309 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56821.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:59:43,306 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8461, 5.9470, 4.9467, 2.6188, 5.0614, 5.6564, 5.4684, 5.1491], + device='cuda:1'), covar=tensor([0.0532, 0.0372, 0.0857, 0.4330, 0.0713, 0.0592, 0.0911, 0.0574], + device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0324, 0.0356, 0.0446, 0.0347, 0.0322, 0.0334, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 04:59:48,621 INFO [train.py:901] (1/4) Epoch 8, batch 250, loss[loss=0.2942, simple_loss=0.3664, pruned_loss=0.1109, over 8246.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3307, pruned_loss=0.09581, over 1163874.54 frames. ], batch size: 24, lr: 9.90e-03, grad_scale: 8.0 +2023-02-06 04:59:51,337 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.703e+02 3.318e+02 4.204e+02 1.022e+03, threshold=6.636e+02, percent-clipped=1.0 +2023-02-06 04:59:56,868 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 05:00:06,243 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 05:00:21,369 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56878.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:00:23,889 INFO [train.py:901] (1/4) Epoch 8, batch 300, loss[loss=0.2738, simple_loss=0.3413, pruned_loss=0.1031, over 8759.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.331, pruned_loss=0.09609, over 1263733.77 frames. ], batch size: 30, lr: 9.89e-03, grad_scale: 8.0 +2023-02-06 05:00:26,003 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56885.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 05:00:28,633 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4117, 1.2272, 1.3352, 1.1038, 0.8252, 1.1552, 1.1400, 1.0546], + device='cuda:1'), covar=tensor([0.0613, 0.1292, 0.1905, 0.1414, 0.0582, 0.1582, 0.0732, 0.0653], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0162, 0.0201, 0.0164, 0.0112, 0.0169, 0.0124, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 05:00:38,052 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56903.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:00:54,955 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56926.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:00:58,777 INFO [train.py:901] (1/4) Epoch 8, batch 350, loss[loss=0.2381, simple_loss=0.303, pruned_loss=0.08664, over 7983.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3317, pruned_loss=0.09692, over 1342846.27 frames. ], batch size: 21, lr: 9.89e-03, grad_scale: 8.0 +2023-02-06 05:01:01,452 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.612e+02 3.168e+02 3.951e+02 1.059e+03, threshold=6.336e+02, percent-clipped=3.0 +2023-02-06 05:01:02,328 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3596, 1.2641, 1.5090, 1.1679, 0.8723, 1.3259, 1.2709, 1.2275], + device='cuda:1'), covar=tensor([0.0588, 0.1296, 0.1800, 0.1404, 0.0564, 0.1555, 0.0658, 0.0575], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0162, 0.0200, 0.0163, 0.0112, 0.0168, 0.0123, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 05:01:33,168 INFO [train.py:901] (1/4) Epoch 8, batch 400, loss[loss=0.2074, simple_loss=0.2915, pruned_loss=0.06168, over 8088.00 frames. ], tot_loss[loss=0.2601, simple_loss=0.3294, pruned_loss=0.09544, over 1405336.86 frames. ], batch size: 21, lr: 9.89e-03, grad_scale: 8.0 +2023-02-06 05:01:45,438 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57000.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 05:01:52,567 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57011.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:01:53,814 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57013.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:02:07,050 INFO [train.py:901] (1/4) Epoch 8, batch 450, loss[loss=0.2883, simple_loss=0.3594, pruned_loss=0.1085, over 8398.00 frames. ], tot_loss[loss=0.2606, simple_loss=0.3295, pruned_loss=0.09588, over 1449130.28 frames. ], batch size: 49, lr: 9.88e-03, grad_scale: 8.0 +2023-02-06 05:02:10,304 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 2.769e+02 3.532e+02 4.551e+02 9.004e+02, threshold=7.064e+02, percent-clipped=7.0 +2023-02-06 05:02:19,152 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3936, 1.9787, 3.2013, 2.5276, 2.6717, 2.0974, 1.5814, 1.3610], + device='cuda:1'), covar=tensor([0.2827, 0.3008, 0.0754, 0.1798, 0.1518, 0.1510, 0.1367, 0.3059], + device='cuda:1'), in_proj_covar=tensor([0.0839, 0.0790, 0.0679, 0.0791, 0.0880, 0.0728, 0.0680, 0.0718], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:02:27,140 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8327, 1.2411, 5.8986, 2.1328, 5.1258, 4.9448, 5.4070, 5.3261], + device='cuda:1'), covar=tensor([0.0333, 0.4596, 0.0322, 0.2830, 0.1019, 0.0689, 0.0414, 0.0447], + device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0532, 0.0494, 0.0479, 0.0544, 0.0456, 0.0455, 0.0509], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 05:02:41,864 INFO [train.py:901] (1/4) Epoch 8, batch 500, loss[loss=0.2311, simple_loss=0.3008, pruned_loss=0.08069, over 7810.00 frames. ], tot_loss[loss=0.2615, simple_loss=0.3303, pruned_loss=0.09633, over 1484855.33 frames. ], batch size: 19, lr: 9.88e-03, grad_scale: 8.0 +2023-02-06 05:02:56,786 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8943, 2.1570, 1.8055, 2.6974, 1.3538, 1.5201, 1.9842, 2.2876], + device='cuda:1'), covar=tensor([0.0877, 0.0838, 0.1243, 0.0467, 0.1156, 0.1644, 0.0911, 0.0742], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0233, 0.0272, 0.0219, 0.0229, 0.0266, 0.0271, 0.0237], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 05:03:15,889 INFO [train.py:901] (1/4) Epoch 8, batch 550, loss[loss=0.2909, simple_loss=0.3662, pruned_loss=0.1079, over 8504.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3294, pruned_loss=0.09575, over 1512233.57 frames. ], batch size: 26, lr: 9.87e-03, grad_scale: 8.0 +2023-02-06 05:03:18,519 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.804e+02 2.761e+02 3.532e+02 4.192e+02 1.400e+03, threshold=7.064e+02, percent-clipped=6.0 +2023-02-06 05:03:39,524 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:03:50,876 INFO [train.py:901] (1/4) Epoch 8, batch 600, loss[loss=0.2455, simple_loss=0.3288, pruned_loss=0.08112, over 8132.00 frames. ], tot_loss[loss=0.2584, simple_loss=0.3277, pruned_loss=0.09451, over 1535467.48 frames. ], batch size: 22, lr: 9.87e-03, grad_scale: 8.0 +2023-02-06 05:04:03,002 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 05:04:06,349 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57204.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:04:13,917 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57215.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:04:25,747 INFO [train.py:901] (1/4) Epoch 8, batch 650, loss[loss=0.3072, simple_loss=0.3677, pruned_loss=0.1234, over 8738.00 frames. ], tot_loss[loss=0.2577, simple_loss=0.3268, pruned_loss=0.09427, over 1553620.25 frames. ], batch size: 39, lr: 9.86e-03, grad_scale: 8.0 +2023-02-06 05:04:28,382 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.822e+02 2.474e+02 3.242e+02 4.284e+02 1.059e+03, threshold=6.484e+02, percent-clipped=6.0 +2023-02-06 05:04:41,995 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57256.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 05:04:46,650 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57262.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:04:51,982 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57270.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:04:59,505 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57280.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:05:00,218 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57281.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 05:05:00,683 INFO [train.py:901] (1/4) Epoch 8, batch 700, loss[loss=0.2086, simple_loss=0.2724, pruned_loss=0.0724, over 7705.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3257, pruned_loss=0.09341, over 1566828.09 frames. ], batch size: 18, lr: 9.86e-03, grad_scale: 8.0 +2023-02-06 05:05:34,910 INFO [train.py:901] (1/4) Epoch 8, batch 750, loss[loss=0.2879, simple_loss=0.3405, pruned_loss=0.1176, over 8081.00 frames. ], tot_loss[loss=0.2564, simple_loss=0.3256, pruned_loss=0.09361, over 1576334.47 frames. ], batch size: 21, lr: 9.86e-03, grad_scale: 8.0 +2023-02-06 05:05:38,340 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.846e+02 3.371e+02 4.091e+02 7.333e+02, threshold=6.742e+02, percent-clipped=1.0 +2023-02-06 05:05:49,702 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 05:05:51,162 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57355.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:05:52,589 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57357.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:05:57,731 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 05:06:09,704 INFO [train.py:901] (1/4) Epoch 8, batch 800, loss[loss=0.2948, simple_loss=0.3544, pruned_loss=0.1176, over 8440.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3255, pruned_loss=0.09334, over 1585283.41 frames. ], batch size: 29, lr: 9.85e-03, grad_scale: 16.0 +2023-02-06 05:06:11,937 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57385.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:06:22,048 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1840, 2.5554, 1.9642, 2.9755, 1.3806, 1.6560, 1.9976, 2.5664], + device='cuda:1'), covar=tensor([0.0745, 0.0718, 0.1144, 0.0451, 0.1352, 0.1557, 0.1154, 0.0720], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0231, 0.0269, 0.0214, 0.0227, 0.0262, 0.0266, 0.0234], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 05:06:33,602 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57416.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:06:44,182 INFO [train.py:901] (1/4) Epoch 8, batch 850, loss[loss=0.218, simple_loss=0.2887, pruned_loss=0.07369, over 7813.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3251, pruned_loss=0.09294, over 1591621.54 frames. ], batch size: 20, lr: 9.85e-03, grad_scale: 16.0 +2023-02-06 05:06:46,897 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.664e+02 3.287e+02 4.255e+02 8.769e+02, threshold=6.575e+02, percent-clipped=4.0 +2023-02-06 05:07:11,112 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57470.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:07:12,464 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:07:17,828 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57480.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:07:19,067 INFO [train.py:901] (1/4) Epoch 8, batch 900, loss[loss=0.2936, simple_loss=0.3632, pruned_loss=0.112, over 8316.00 frames. ], tot_loss[loss=0.2567, simple_loss=0.3264, pruned_loss=0.09352, over 1598955.42 frames. ], batch size: 25, lr: 9.84e-03, grad_scale: 16.0 +2023-02-06 05:07:40,918 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57513.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:07:53,666 INFO [train.py:901] (1/4) Epoch 8, batch 950, loss[loss=0.3255, simple_loss=0.3788, pruned_loss=0.1361, over 8248.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3265, pruned_loss=0.0933, over 1604869.40 frames. ], batch size: 24, lr: 9.84e-03, grad_scale: 16.0 +2023-02-06 05:07:56,421 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.713e+02 3.197e+02 4.416e+02 7.629e+02, threshold=6.394e+02, percent-clipped=6.0 +2023-02-06 05:07:56,665 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57536.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:07:56,700 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2612, 1.7764, 2.9695, 2.3118, 2.5173, 2.0114, 1.6451, 1.1336], + device='cuda:1'), covar=tensor([0.2893, 0.3295, 0.0762, 0.1876, 0.1453, 0.1702, 0.1416, 0.3341], + device='cuda:1'), in_proj_covar=tensor([0.0827, 0.0779, 0.0675, 0.0778, 0.0863, 0.0720, 0.0667, 0.0706], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:08:04,584 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57548.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:08:13,024 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57559.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:08:14,317 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 05:08:14,516 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57561.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:08:24,649 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3251, 1.9607, 2.8895, 2.3899, 2.6384, 2.0683, 1.8483, 1.7716], + device='cuda:1'), covar=tensor([0.2033, 0.2506, 0.0609, 0.1376, 0.1085, 0.1353, 0.1115, 0.2233], + device='cuda:1'), in_proj_covar=tensor([0.0827, 0.0779, 0.0677, 0.0778, 0.0864, 0.0722, 0.0670, 0.0709], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:08:29,126 INFO [train.py:901] (1/4) Epoch 8, batch 1000, loss[loss=0.1838, simple_loss=0.2658, pruned_loss=0.05091, over 7543.00 frames. ], tot_loss[loss=0.2567, simple_loss=0.3268, pruned_loss=0.09327, over 1605519.00 frames. ], batch size: 18, lr: 9.83e-03, grad_scale: 16.0 +2023-02-06 05:08:45,898 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57606.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:08:47,892 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 05:09:00,599 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 05:09:03,903 INFO [train.py:901] (1/4) Epoch 8, batch 1050, loss[loss=0.2422, simple_loss=0.3245, pruned_loss=0.08, over 8338.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.3275, pruned_loss=0.09341, over 1608075.03 frames. ], batch size: 25, lr: 9.83e-03, grad_scale: 16.0 +2023-02-06 05:09:06,634 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.778e+02 2.733e+02 3.382e+02 4.210e+02 1.523e+03, threshold=6.765e+02, percent-clipped=11.0 +2023-02-06 05:09:10,051 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57641.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:09:22,119 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 05:09:24,434 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57663.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:09:26,522 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57666.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:09:26,667 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 05:09:31,687 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57674.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:09:37,592 INFO [train.py:901] (1/4) Epoch 8, batch 1100, loss[loss=0.233, simple_loss=0.2957, pruned_loss=0.08519, over 7667.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.3284, pruned_loss=0.09448, over 1609042.36 frames. ], batch size: 19, lr: 9.83e-03, grad_scale: 16.0 +2023-02-06 05:09:57,897 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2034, 2.6251, 2.8984, 1.2772, 3.2665, 1.8926, 1.5058, 1.7441], + device='cuda:1'), covar=tensor([0.0391, 0.0164, 0.0158, 0.0348, 0.0226, 0.0495, 0.0508, 0.0278], + device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0277, 0.0226, 0.0334, 0.0271, 0.0420, 0.0328, 0.0308], + device='cuda:1'), out_proj_covar=tensor([1.1122e-04, 8.4474e-05, 6.8003e-05, 1.0109e-04, 8.3864e-05, 1.3921e-04, + 1.0172e-04, 9.4731e-05], device='cuda:1') +2023-02-06 05:10:05,399 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57721.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:08,897 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57726.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:10,255 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57728.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:10,705 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 05:10:12,775 INFO [train.py:901] (1/4) Epoch 8, batch 1150, loss[loss=0.3028, simple_loss=0.3814, pruned_loss=0.1121, over 8596.00 frames. ], tot_loss[loss=0.2573, simple_loss=0.3275, pruned_loss=0.0935, over 1614681.61 frames. ], batch size: 39, lr: 9.82e-03, grad_scale: 16.0 +2023-02-06 05:10:15,551 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.752e+02 3.349e+02 4.211e+02 1.172e+03, threshold=6.698e+02, percent-clipped=4.0 +2023-02-06 05:10:21,932 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9907, 1.4463, 6.0883, 1.8630, 5.3924, 5.2766, 5.7844, 5.6051], + device='cuda:1'), covar=tensor([0.0514, 0.3941, 0.0294, 0.2952, 0.0985, 0.0584, 0.0383, 0.0410], + device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0527, 0.0491, 0.0473, 0.0534, 0.0449, 0.0447, 0.0507], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 05:10:25,409 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2431, 2.5277, 1.7553, 2.0666, 1.8897, 1.3222, 1.7332, 2.0318], + device='cuda:1'), covar=tensor([0.1312, 0.0302, 0.0985, 0.0526, 0.0693, 0.1309, 0.0865, 0.0725], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0234, 0.0313, 0.0300, 0.0306, 0.0314, 0.0334, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 05:10:26,728 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57751.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:26,795 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57751.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:28,154 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57753.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:32,811 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:47,638 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2150, 1.6568, 1.6002, 1.3299, 1.1495, 1.4752, 1.7557, 1.8427], + device='cuda:1'), covar=tensor([0.0506, 0.1133, 0.1752, 0.1326, 0.0567, 0.1536, 0.0665, 0.0563], + device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0163, 0.0200, 0.0165, 0.0111, 0.0170, 0.0124, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 05:10:48,150 INFO [train.py:901] (1/4) Epoch 8, batch 1200, loss[loss=0.2285, simple_loss=0.2819, pruned_loss=0.08755, over 7534.00 frames. ], tot_loss[loss=0.2567, simple_loss=0.3268, pruned_loss=0.09334, over 1614086.27 frames. ], batch size: 18, lr: 9.82e-03, grad_scale: 16.0 +2023-02-06 05:10:58,922 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9577, 1.6460, 4.1032, 1.4924, 2.1999, 4.5917, 4.8219, 3.4937], + device='cuda:1'), covar=tensor([0.1710, 0.2028, 0.0383, 0.2707, 0.1294, 0.0464, 0.0426, 0.1271], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0281, 0.0239, 0.0272, 0.0247, 0.0218, 0.0288, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 05:11:17,882 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:11:23,251 INFO [train.py:901] (1/4) Epoch 8, batch 1250, loss[loss=0.2675, simple_loss=0.3371, pruned_loss=0.09897, over 8493.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.3272, pruned_loss=0.09387, over 1612741.18 frames. ], batch size: 28, lr: 9.81e-03, grad_scale: 16.0 +2023-02-06 05:11:25,903 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.817e+02 3.577e+02 4.191e+02 8.690e+02, threshold=7.155e+02, percent-clipped=5.0 +2023-02-06 05:11:41,388 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57857.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:11:53,673 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:11:58,371 INFO [train.py:901] (1/4) Epoch 8, batch 1300, loss[loss=0.2733, simple_loss=0.343, pruned_loss=0.1018, over 8187.00 frames. ], tot_loss[loss=0.258, simple_loss=0.3279, pruned_loss=0.09407, over 1615852.12 frames. ], batch size: 23, lr: 9.81e-03, grad_scale: 16.0 +2023-02-06 05:12:24,246 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57919.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:12:31,597 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2129, 1.9715, 3.0434, 2.4263, 2.5667, 2.0821, 1.6742, 1.2848], + device='cuda:1'), covar=tensor([0.3282, 0.3189, 0.0805, 0.1778, 0.1551, 0.1700, 0.1475, 0.3284], + device='cuda:1'), in_proj_covar=tensor([0.0836, 0.0784, 0.0676, 0.0779, 0.0863, 0.0720, 0.0666, 0.0709], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:12:32,236 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57930.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:12:33,380 INFO [train.py:901] (1/4) Epoch 8, batch 1350, loss[loss=0.2111, simple_loss=0.2981, pruned_loss=0.06211, over 8251.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.3278, pruned_loss=0.09418, over 1614842.48 frames. ], batch size: 22, lr: 9.80e-03, grad_scale: 16.0 +2023-02-06 05:12:35,571 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4516, 4.5716, 4.0757, 2.1045, 3.9808, 4.0975, 4.2307, 3.6214], + device='cuda:1'), covar=tensor([0.0767, 0.0434, 0.0871, 0.4149, 0.0723, 0.0817, 0.1050, 0.0758], + device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0325, 0.0359, 0.0447, 0.0353, 0.0329, 0.0335, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:12:36,127 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 2.787e+02 3.281e+02 4.089e+02 1.129e+03, threshold=6.562e+02, percent-clipped=4.0 +2023-02-06 05:12:38,227 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:12:41,653 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:12:49,718 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57955.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:13:01,727 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57972.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:13:05,179 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57977.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:13:08,423 INFO [train.py:901] (1/4) Epoch 8, batch 1400, loss[loss=0.2369, simple_loss=0.321, pruned_loss=0.07638, over 8362.00 frames. ], tot_loss[loss=0.2591, simple_loss=0.3286, pruned_loss=0.0948, over 1617763.48 frames. ], batch size: 24, lr: 9.80e-03, grad_scale: 16.0 +2023-02-06 05:13:23,094 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58002.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:13:41,319 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 05:13:42,543 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 05:13:43,360 INFO [train.py:901] (1/4) Epoch 8, batch 1450, loss[loss=0.2725, simple_loss=0.3411, pruned_loss=0.1019, over 7808.00 frames. ], tot_loss[loss=0.2593, simple_loss=0.3295, pruned_loss=0.09454, over 1621475.17 frames. ], batch size: 20, lr: 9.80e-03, grad_scale: 16.0 +2023-02-06 05:13:46,043 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.668e+02 3.298e+02 4.223e+02 1.032e+03, threshold=6.596e+02, percent-clipped=5.0 +2023-02-06 05:14:18,669 INFO [train.py:901] (1/4) Epoch 8, batch 1500, loss[loss=0.2781, simple_loss=0.3421, pruned_loss=0.107, over 8584.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3305, pruned_loss=0.09545, over 1623204.60 frames. ], batch size: 34, lr: 9.79e-03, grad_scale: 16.0 +2023-02-06 05:14:28,078 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58095.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:14:31,216 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 05:14:52,815 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58131.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:14:53,267 INFO [train.py:901] (1/4) Epoch 8, batch 1550, loss[loss=0.2386, simple_loss=0.3163, pruned_loss=0.08043, over 8601.00 frames. ], tot_loss[loss=0.259, simple_loss=0.3288, pruned_loss=0.09463, over 1618611.77 frames. ], batch size: 49, lr: 9.79e-03, grad_scale: 16.0 +2023-02-06 05:14:56,006 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.601e+02 3.218e+02 3.979e+02 6.246e+02, threshold=6.435e+02, percent-clipped=0.0 +2023-02-06 05:15:10,024 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58156.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:15:12,036 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58159.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:15:14,756 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58162.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:15:27,769 INFO [train.py:901] (1/4) Epoch 8, batch 1600, loss[loss=0.2996, simple_loss=0.3626, pruned_loss=0.1183, over 8357.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3299, pruned_loss=0.0958, over 1616443.21 frames. ], batch size: 24, lr: 9.78e-03, grad_scale: 16.0 +2023-02-06 05:15:37,317 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58195.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:15:47,383 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58210.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:15:54,054 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58220.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:16:00,056 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58228.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:16:02,612 INFO [train.py:901] (1/4) Epoch 8, batch 1650, loss[loss=0.3217, simple_loss=0.3667, pruned_loss=0.1383, over 6781.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3298, pruned_loss=0.09561, over 1615228.60 frames. ], batch size: 71, lr: 9.78e-03, grad_scale: 16.0 +2023-02-06 05:16:05,269 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.785e+02 3.241e+02 4.331e+02 1.468e+03, threshold=6.482e+02, percent-clipped=4.0 +2023-02-06 05:16:16,840 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58253.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:16:37,572 INFO [train.py:901] (1/4) Epoch 8, batch 1700, loss[loss=0.2357, simple_loss=0.3072, pruned_loss=0.08207, over 7963.00 frames. ], tot_loss[loss=0.2593, simple_loss=0.3286, pruned_loss=0.09496, over 1614127.54 frames. ], batch size: 21, lr: 9.78e-03, grad_scale: 16.0 +2023-02-06 05:17:00,580 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.4574, 5.4887, 4.9113, 2.0451, 4.9321, 5.1986, 5.2038, 4.7307], + device='cuda:1'), covar=tensor([0.0574, 0.0462, 0.0828, 0.4630, 0.0637, 0.0593, 0.1002, 0.0571], + device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0324, 0.0355, 0.0444, 0.0349, 0.0325, 0.0335, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:17:11,722 INFO [train.py:901] (1/4) Epoch 8, batch 1750, loss[loss=0.2555, simple_loss=0.3214, pruned_loss=0.09482, over 8239.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.3277, pruned_loss=0.09365, over 1614929.50 frames. ], batch size: 22, lr: 9.77e-03, grad_scale: 16.0 +2023-02-06 05:17:15,040 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.673e+02 2.752e+02 3.204e+02 3.949e+02 8.384e+02, threshold=6.409e+02, percent-clipped=4.0 +2023-02-06 05:17:41,591 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-06 05:17:43,449 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-06 05:17:45,746 INFO [train.py:901] (1/4) Epoch 8, batch 1800, loss[loss=0.2463, simple_loss=0.3245, pruned_loss=0.08402, over 8526.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3265, pruned_loss=0.092, over 1617845.19 frames. ], batch size: 31, lr: 9.77e-03, grad_scale: 16.0 +2023-02-06 05:18:06,234 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 05:18:21,312 INFO [train.py:901] (1/4) Epoch 8, batch 1850, loss[loss=0.2536, simple_loss=0.3259, pruned_loss=0.09065, over 8191.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3267, pruned_loss=0.0921, over 1620875.01 frames. ], batch size: 23, lr: 9.76e-03, grad_scale: 16.0 +2023-02-06 05:18:24,004 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.956e+02 3.603e+02 4.636e+02 8.044e+02, threshold=7.207e+02, percent-clipped=5.0 +2023-02-06 05:18:45,039 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58466.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:18:55,854 INFO [train.py:901] (1/4) Epoch 8, batch 1900, loss[loss=0.2627, simple_loss=0.3361, pruned_loss=0.0946, over 8348.00 frames. ], tot_loss[loss=0.255, simple_loss=0.3258, pruned_loss=0.09205, over 1619216.29 frames. ], batch size: 26, lr: 9.76e-03, grad_scale: 16.0 +2023-02-06 05:19:01,998 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58491.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:19:10,026 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58503.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:19:12,313 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 +2023-02-06 05:19:12,698 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58506.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:19:19,280 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 05:19:30,166 INFO [train.py:901] (1/4) Epoch 8, batch 1950, loss[loss=0.2475, simple_loss=0.3207, pruned_loss=0.08719, over 7641.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3261, pruned_loss=0.09253, over 1619102.58 frames. ], batch size: 19, lr: 9.75e-03, grad_scale: 16.0 +2023-02-06 05:19:30,817 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 05:19:32,606 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-02-06 05:19:32,812 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.699e+02 3.417e+02 4.103e+02 8.210e+02, threshold=6.834e+02, percent-clipped=5.0 +2023-02-06 05:19:50,870 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 05:20:04,902 INFO [train.py:901] (1/4) Epoch 8, batch 2000, loss[loss=0.2649, simple_loss=0.3409, pruned_loss=0.0944, over 8638.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3257, pruned_loss=0.09211, over 1617949.02 frames. ], batch size: 34, lr: 9.75e-03, grad_scale: 8.0 +2023-02-06 05:20:29,405 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58618.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:20:31,398 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58621.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:20:36,253 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-02-06 05:20:39,664 INFO [train.py:901] (1/4) Epoch 8, batch 2050, loss[loss=0.3133, simple_loss=0.3645, pruned_loss=0.1311, over 8772.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3263, pruned_loss=0.0925, over 1620260.53 frames. ], batch size: 30, lr: 9.75e-03, grad_scale: 8.0 +2023-02-06 05:20:42,943 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.764e+02 2.785e+02 3.396e+02 4.687e+02 1.585e+03, threshold=6.792e+02, percent-clipped=4.0 +2023-02-06 05:21:13,661 INFO [train.py:901] (1/4) Epoch 8, batch 2100, loss[loss=0.2375, simple_loss=0.314, pruned_loss=0.0805, over 8181.00 frames. ], tot_loss[loss=0.2565, simple_loss=0.3268, pruned_loss=0.09308, over 1617681.92 frames. ], batch size: 23, lr: 9.74e-03, grad_scale: 8.0 +2023-02-06 05:21:25,816 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58699.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:21:47,831 INFO [train.py:901] (1/4) Epoch 8, batch 2150, loss[loss=0.2256, simple_loss=0.2939, pruned_loss=0.07868, over 7688.00 frames. ], tot_loss[loss=0.2529, simple_loss=0.3231, pruned_loss=0.09131, over 1609460.00 frames. ], batch size: 18, lr: 9.74e-03, grad_scale: 8.0 +2023-02-06 05:21:51,088 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.839e+02 2.818e+02 3.372e+02 4.104e+02 8.704e+02, threshold=6.743e+02, percent-clipped=2.0 +2023-02-06 05:21:51,247 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0885, 1.2744, 4.2258, 1.5017, 3.6560, 3.4415, 3.7754, 3.6303], + device='cuda:1'), covar=tensor([0.0449, 0.4195, 0.0464, 0.3239, 0.1056, 0.0755, 0.0480, 0.0629], + device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0528, 0.0503, 0.0473, 0.0540, 0.0457, 0.0454, 0.0510], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 05:22:23,675 INFO [train.py:901] (1/4) Epoch 8, batch 2200, loss[loss=0.3244, simple_loss=0.3677, pruned_loss=0.1405, over 7053.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3247, pruned_loss=0.09237, over 1608630.81 frames. ], batch size: 72, lr: 9.73e-03, grad_scale: 8.0 +2023-02-06 05:22:31,531 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58793.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:22:49,412 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5539, 2.8008, 1.8141, 2.0173, 2.3241, 1.5671, 1.9407, 2.1964], + device='cuda:1'), covar=tensor([0.1276, 0.0277, 0.0872, 0.0621, 0.0567, 0.1069, 0.0880, 0.0749], + device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0231, 0.0308, 0.0294, 0.0300, 0.0313, 0.0334, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 05:22:53,124 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 05:22:58,789 INFO [train.py:901] (1/4) Epoch 8, batch 2250, loss[loss=0.229, simple_loss=0.3168, pruned_loss=0.07059, over 8322.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3262, pruned_loss=0.0928, over 1614787.03 frames. ], batch size: 26, lr: 9.73e-03, grad_scale: 8.0 +2023-02-06 05:23:02,309 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.890e+02 2.600e+02 3.138e+02 4.259e+02 8.800e+02, threshold=6.276e+02, percent-clipped=5.0 +2023-02-06 05:23:08,952 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-02-06 05:23:29,266 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58874.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:23:31,271 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58877.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:23:34,472 INFO [train.py:901] (1/4) Epoch 8, batch 2300, loss[loss=0.218, simple_loss=0.3018, pruned_loss=0.06705, over 8028.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3253, pruned_loss=0.09277, over 1611467.89 frames. ], batch size: 22, lr: 9.73e-03, grad_scale: 8.0 +2023-02-06 05:23:46,510 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58899.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:23:48,578 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58902.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:24:03,549 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58924.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:24:09,482 INFO [train.py:901] (1/4) Epoch 8, batch 2350, loss[loss=0.2645, simple_loss=0.3447, pruned_loss=0.09215, over 8249.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.3263, pruned_loss=0.09306, over 1614269.57 frames. ], batch size: 24, lr: 9.72e-03, grad_scale: 8.0 +2023-02-06 05:24:12,942 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.611e+02 3.221e+02 3.780e+02 8.999e+02, threshold=6.441e+02, percent-clipped=2.0 +2023-02-06 05:24:44,047 INFO [train.py:901] (1/4) Epoch 8, batch 2400, loss[loss=0.2404, simple_loss=0.3014, pruned_loss=0.08966, over 7555.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3257, pruned_loss=0.09267, over 1612041.59 frames. ], batch size: 18, lr: 9.72e-03, grad_scale: 8.0 +2023-02-06 05:25:18,661 INFO [train.py:901] (1/4) Epoch 8, batch 2450, loss[loss=0.3225, simple_loss=0.384, pruned_loss=0.1305, over 8441.00 frames. ], tot_loss[loss=0.256, simple_loss=0.3264, pruned_loss=0.09278, over 1613618.61 frames. ], batch size: 29, lr: 9.71e-03, grad_scale: 8.0 +2023-02-06 05:25:21,882 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 3.001e+02 3.706e+02 4.542e+02 9.599e+02, threshold=7.413e+02, percent-clipped=3.0 +2023-02-06 05:25:26,104 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:25:52,508 INFO [train.py:901] (1/4) Epoch 8, batch 2500, loss[loss=0.273, simple_loss=0.3547, pruned_loss=0.09561, over 8097.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3257, pruned_loss=0.09259, over 1614462.97 frames. ], batch size: 23, lr: 9.71e-03, grad_scale: 8.0 +2023-02-06 05:26:27,546 INFO [train.py:901] (1/4) Epoch 8, batch 2550, loss[loss=0.2879, simple_loss=0.3529, pruned_loss=0.1114, over 8606.00 frames. ], tot_loss[loss=0.256, simple_loss=0.3255, pruned_loss=0.09331, over 1614076.47 frames. ], batch size: 34, lr: 9.71e-03, grad_scale: 8.0 +2023-02-06 05:26:29,740 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0434, 1.4516, 1.5075, 1.3194, 1.1150, 1.3779, 1.4985, 1.6617], + device='cuda:1'), covar=tensor([0.0560, 0.1280, 0.1835, 0.1431, 0.0619, 0.1555, 0.0739, 0.0568], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0161, 0.0200, 0.0165, 0.0111, 0.0170, 0.0124, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 05:26:30,876 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.676e+02 3.180e+02 4.175e+02 9.807e+02, threshold=6.360e+02, percent-clipped=4.0 +2023-02-06 05:26:30,966 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59137.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:26:45,980 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59158.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:27:03,045 INFO [train.py:901] (1/4) Epoch 8, batch 2600, loss[loss=0.2363, simple_loss=0.3092, pruned_loss=0.08172, over 7651.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3248, pruned_loss=0.09276, over 1609227.47 frames. ], batch size: 19, lr: 9.70e-03, grad_scale: 8.0 +2023-02-06 05:27:03,179 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59182.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:27:18,803 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4437, 1.4017, 4.4375, 1.7943, 2.4166, 5.1297, 4.9981, 4.4865], + device='cuda:1'), covar=tensor([0.1037, 0.1596, 0.0230, 0.1916, 0.0983, 0.0175, 0.0292, 0.0466], + device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0285, 0.0243, 0.0281, 0.0249, 0.0226, 0.0296, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 05:27:38,220 INFO [train.py:901] (1/4) Epoch 8, batch 2650, loss[loss=0.2552, simple_loss=0.332, pruned_loss=0.0892, over 8467.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3255, pruned_loss=0.09291, over 1612248.52 frames. ], batch size: 29, lr: 9.70e-03, grad_scale: 8.0 +2023-02-06 05:27:41,643 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.757e+02 3.213e+02 4.207e+02 1.360e+03, threshold=6.426e+02, percent-clipped=6.0 +2023-02-06 05:27:52,003 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59252.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:28:03,086 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59268.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:28:12,228 INFO [train.py:901] (1/4) Epoch 8, batch 2700, loss[loss=0.267, simple_loss=0.3086, pruned_loss=0.1126, over 7922.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3244, pruned_loss=0.09236, over 1614189.48 frames. ], batch size: 20, lr: 9.69e-03, grad_scale: 8.0 +2023-02-06 05:28:46,767 INFO [train.py:901] (1/4) Epoch 8, batch 2750, loss[loss=0.2228, simple_loss=0.3069, pruned_loss=0.06929, over 8242.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.3243, pruned_loss=0.09201, over 1616411.07 frames. ], batch size: 22, lr: 9.69e-03, grad_scale: 8.0 +2023-02-06 05:28:50,099 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.941e+02 2.846e+02 3.367e+02 4.274e+02 9.837e+02, threshold=6.735e+02, percent-clipped=6.0 +2023-02-06 05:29:14,891 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-02-06 05:29:22,662 INFO [train.py:901] (1/4) Epoch 8, batch 2800, loss[loss=0.2864, simple_loss=0.346, pruned_loss=0.1134, over 7974.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3244, pruned_loss=0.09152, over 1619507.72 frames. ], batch size: 21, lr: 9.69e-03, grad_scale: 8.0 +2023-02-06 05:29:23,476 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59383.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:29:44,991 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59414.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:29:50,442 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 05:29:56,663 INFO [train.py:901] (1/4) Epoch 8, batch 2850, loss[loss=0.2761, simple_loss=0.3336, pruned_loss=0.1093, over 8082.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3252, pruned_loss=0.09181, over 1619772.21 frames. ], batch size: 21, lr: 9.68e-03, grad_scale: 8.0 +2023-02-06 05:30:00,150 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.577e+02 2.974e+02 3.773e+02 5.956e+02, threshold=5.948e+02, percent-clipped=0.0 +2023-02-06 05:30:01,814 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59439.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:30:16,826 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7569, 1.2045, 3.9075, 1.4304, 3.4596, 3.2136, 3.4723, 3.4202], + device='cuda:1'), covar=tensor([0.0528, 0.3996, 0.0478, 0.2867, 0.1053, 0.0777, 0.0582, 0.0577], + device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0531, 0.0508, 0.0473, 0.0549, 0.0460, 0.0458, 0.0508], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 05:30:32,530 INFO [train.py:901] (1/4) Epoch 8, batch 2900, loss[loss=0.257, simple_loss=0.3286, pruned_loss=0.09266, over 8332.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3254, pruned_loss=0.0922, over 1616842.37 frames. ], batch size: 26, lr: 9.68e-03, grad_scale: 8.0 +2023-02-06 05:30:39,660 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6776, 2.2308, 3.6604, 2.8308, 2.9113, 2.2752, 1.9138, 1.9038], + device='cuda:1'), covar=tensor([0.2888, 0.3383, 0.0871, 0.2131, 0.1836, 0.1750, 0.1384, 0.3583], + device='cuda:1'), in_proj_covar=tensor([0.0844, 0.0801, 0.0685, 0.0787, 0.0885, 0.0740, 0.0673, 0.0722], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:30:47,128 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2107, 1.1474, 1.1735, 1.1754, 0.8857, 1.2897, 0.0563, 0.9027], + device='cuda:1'), covar=tensor([0.2616, 0.1867, 0.0750, 0.1409, 0.4406, 0.0771, 0.3822, 0.1805], + device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0146, 0.0083, 0.0193, 0.0230, 0.0088, 0.0151, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:30:51,219 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59508.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:30:59,190 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:31:03,060 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59526.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:31:04,349 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 05:31:07,543 INFO [train.py:901] (1/4) Epoch 8, batch 2950, loss[loss=0.2898, simple_loss=0.3655, pruned_loss=0.1071, over 8023.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3269, pruned_loss=0.09351, over 1611746.91 frames. ], batch size: 22, lr: 9.67e-03, grad_scale: 8.0 +2023-02-06 05:31:08,345 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:31:10,778 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.890e+02 2.847e+02 3.468e+02 5.057e+02 9.591e+02, threshold=6.936e+02, percent-clipped=13.0 +2023-02-06 05:31:42,183 INFO [train.py:901] (1/4) Epoch 8, batch 3000, loss[loss=0.2534, simple_loss=0.3279, pruned_loss=0.0894, over 8350.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3256, pruned_loss=0.09289, over 1610391.66 frames. ], batch size: 24, lr: 9.67e-03, grad_scale: 8.0 +2023-02-06 05:31:42,183 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 05:31:54,427 INFO [train.py:935] (1/4) Epoch 8, validation: loss=0.2021, simple_loss=0.3001, pruned_loss=0.05199, over 944034.00 frames. +2023-02-06 05:31:54,428 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 05:32:30,902 INFO [train.py:901] (1/4) Epoch 8, batch 3050, loss[loss=0.2656, simple_loss=0.3376, pruned_loss=0.09677, over 8463.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3248, pruned_loss=0.09286, over 1608805.61 frames. ], batch size: 25, lr: 9.67e-03, grad_scale: 8.0 +2023-02-06 05:32:34,251 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.623e+02 3.324e+02 4.059e+02 7.396e+02, threshold=6.648e+02, percent-clipped=1.0 +2023-02-06 05:32:35,816 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59639.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:32:37,140 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59641.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:32:52,395 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59664.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:33:05,113 INFO [train.py:901] (1/4) Epoch 8, batch 3100, loss[loss=0.2254, simple_loss=0.311, pruned_loss=0.06994, over 8293.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3256, pruned_loss=0.09327, over 1611145.45 frames. ], batch size: 23, lr: 9.66e-03, grad_scale: 8.0 +2023-02-06 05:33:40,028 INFO [train.py:901] (1/4) Epoch 8, batch 3150, loss[loss=0.2443, simple_loss=0.309, pruned_loss=0.08983, over 7929.00 frames. ], tot_loss[loss=0.2567, simple_loss=0.326, pruned_loss=0.09372, over 1611396.61 frames. ], batch size: 20, lr: 9.66e-03, grad_scale: 8.0 +2023-02-06 05:33:43,232 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 2.894e+02 3.427e+02 4.526e+02 8.691e+02, threshold=6.853e+02, percent-clipped=4.0 +2023-02-06 05:33:57,517 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 05:34:14,631 INFO [train.py:901] (1/4) Epoch 8, batch 3200, loss[loss=0.2881, simple_loss=0.3635, pruned_loss=0.1064, over 8502.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.325, pruned_loss=0.09297, over 1613003.08 frames. ], batch size: 49, lr: 9.65e-03, grad_scale: 8.0 +2023-02-06 05:34:23,260 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 05:34:50,973 INFO [train.py:901] (1/4) Epoch 8, batch 3250, loss[loss=0.249, simple_loss=0.3113, pruned_loss=0.09337, over 7237.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3263, pruned_loss=0.09317, over 1615066.67 frames. ], batch size: 16, lr: 9.65e-03, grad_scale: 8.0 +2023-02-06 05:34:54,318 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.545e+02 3.201e+02 4.295e+02 9.179e+02, threshold=6.402e+02, percent-clipped=6.0 +2023-02-06 05:35:13,090 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59864.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:14,536 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59866.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:25,226 INFO [train.py:901] (1/4) Epoch 8, batch 3300, loss[loss=0.2055, simple_loss=0.2755, pruned_loss=0.06776, over 7425.00 frames. ], tot_loss[loss=0.2565, simple_loss=0.3261, pruned_loss=0.0935, over 1613899.33 frames. ], batch size: 17, lr: 9.65e-03, grad_scale: 8.0 +2023-02-06 05:35:35,299 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59897.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:37,375 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.95 vs. limit=5.0 +2023-02-06 05:35:41,750 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59907.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:52,438 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59922.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:53,730 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4124, 1.5372, 1.4383, 1.4053, 1.1613, 1.3909, 1.8922, 1.7940], + device='cuda:1'), covar=tensor([0.0517, 0.1183, 0.1803, 0.1395, 0.0564, 0.1555, 0.0672, 0.0575], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0160, 0.0199, 0.0164, 0.0109, 0.0169, 0.0123, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 05:35:59,535 INFO [train.py:901] (1/4) Epoch 8, batch 3350, loss[loss=0.233, simple_loss=0.2905, pruned_loss=0.08778, over 7194.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.327, pruned_loss=0.09396, over 1611667.55 frames. ], batch size: 16, lr: 9.64e-03, grad_scale: 8.0 +2023-02-06 05:36:02,906 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.805e+02 2.795e+02 3.400e+02 4.166e+02 8.824e+02, threshold=6.801e+02, percent-clipped=5.0 +2023-02-06 05:36:25,158 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2557, 1.6151, 1.4863, 1.4000, 1.0742, 1.3132, 1.6502, 1.5780], + device='cuda:1'), covar=tensor([0.0486, 0.1191, 0.1744, 0.1352, 0.0583, 0.1561, 0.0684, 0.0597], + device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0162, 0.0200, 0.0165, 0.0111, 0.0170, 0.0124, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 05:36:31,821 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:36:33,704 INFO [train.py:901] (1/4) Epoch 8, batch 3400, loss[loss=0.303, simple_loss=0.3595, pruned_loss=0.1232, over 8495.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3256, pruned_loss=0.09298, over 1609988.98 frames. ], batch size: 26, lr: 9.64e-03, grad_scale: 8.0 +2023-02-06 05:36:51,164 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 05:37:08,461 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9377, 1.5813, 2.2502, 1.8914, 2.0038, 1.7955, 1.4742, 0.6673], + device='cuda:1'), covar=tensor([0.3062, 0.2965, 0.0900, 0.1718, 0.1381, 0.1714, 0.1469, 0.2833], + device='cuda:1'), in_proj_covar=tensor([0.0839, 0.0805, 0.0683, 0.0784, 0.0886, 0.0735, 0.0675, 0.0719], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:37:09,550 INFO [train.py:901] (1/4) Epoch 8, batch 3450, loss[loss=0.2278, simple_loss=0.2941, pruned_loss=0.08072, over 5963.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.3248, pruned_loss=0.09242, over 1608669.15 frames. ], batch size: 13, lr: 9.63e-03, grad_scale: 8.0 +2023-02-06 05:37:12,877 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.668e+02 3.106e+02 3.891e+02 9.201e+02, threshold=6.211e+02, percent-clipped=2.0 +2023-02-06 05:37:27,911 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60058.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:37:36,417 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3219, 1.2723, 4.5358, 1.7373, 4.0008, 3.7901, 4.0830, 3.9459], + device='cuda:1'), covar=tensor([0.0489, 0.3729, 0.0385, 0.2770, 0.1004, 0.0745, 0.0460, 0.0526], + device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0532, 0.0512, 0.0474, 0.0549, 0.0463, 0.0458, 0.0509], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 05:37:36,426 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60070.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:37:44,253 INFO [train.py:901] (1/4) Epoch 8, batch 3500, loss[loss=0.2289, simple_loss=0.2962, pruned_loss=0.08081, over 7532.00 frames. ], tot_loss[loss=0.254, simple_loss=0.324, pruned_loss=0.09198, over 1609181.63 frames. ], batch size: 18, lr: 9.63e-03, grad_scale: 8.0 +2023-02-06 05:37:45,065 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60083.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:37:53,893 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60096.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:38:02,913 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 05:38:18,648 INFO [train.py:901] (1/4) Epoch 8, batch 3550, loss[loss=0.2245, simple_loss=0.2925, pruned_loss=0.07825, over 7231.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.3243, pruned_loss=0.09207, over 1609707.60 frames. ], batch size: 16, lr: 9.63e-03, grad_scale: 8.0 +2023-02-06 05:38:22,097 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.844e+02 3.449e+02 4.512e+02 7.529e+02, threshold=6.898e+02, percent-clipped=5.0 +2023-02-06 05:38:43,014 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-02-06 05:38:54,366 INFO [train.py:901] (1/4) Epoch 8, batch 3600, loss[loss=0.2532, simple_loss=0.3357, pruned_loss=0.08529, over 8522.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.325, pruned_loss=0.09258, over 1606652.26 frames. ], batch size: 28, lr: 9.62e-03, grad_scale: 8.0 +2023-02-06 05:39:14,697 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60210.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:39:18,777 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60216.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:39:29,810 INFO [train.py:901] (1/4) Epoch 8, batch 3650, loss[loss=0.2948, simple_loss=0.3565, pruned_loss=0.1166, over 8039.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3251, pruned_loss=0.09282, over 1608888.88 frames. ], batch size: 22, lr: 9.62e-03, grad_scale: 8.0 +2023-02-06 05:39:32,067 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60235.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:39:33,129 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.702e+02 3.457e+02 4.155e+02 9.631e+02, threshold=6.915e+02, percent-clipped=4.0 +2023-02-06 05:39:42,738 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60251.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:39:48,807 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60260.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:40:02,757 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 05:40:04,766 INFO [train.py:901] (1/4) Epoch 8, batch 3700, loss[loss=0.231, simple_loss=0.2909, pruned_loss=0.08554, over 7555.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3257, pruned_loss=0.09279, over 1610873.10 frames. ], batch size: 18, lr: 9.61e-03, grad_scale: 8.0 +2023-02-06 05:40:34,567 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60325.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:40:39,110 INFO [train.py:901] (1/4) Epoch 8, batch 3750, loss[loss=0.2695, simple_loss=0.3477, pruned_loss=0.09562, over 8463.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.3248, pruned_loss=0.09184, over 1613258.92 frames. ], batch size: 27, lr: 9.61e-03, grad_scale: 8.0 +2023-02-06 05:40:43,054 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.685e+02 3.295e+02 3.882e+02 8.274e+02, threshold=6.589e+02, percent-clipped=2.0 +2023-02-06 05:41:02,724 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:41:03,425 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4722, 1.8644, 1.9685, 1.2423, 2.0752, 1.4954, 0.5076, 1.6466], + device='cuda:1'), covar=tensor([0.0352, 0.0201, 0.0131, 0.0251, 0.0227, 0.0481, 0.0492, 0.0155], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0279, 0.0228, 0.0335, 0.0267, 0.0424, 0.0330, 0.0307], + device='cuda:1'), out_proj_covar=tensor([1.0899e-04, 8.3723e-05, 6.7981e-05, 1.0004e-04, 8.1914e-05, 1.3918e-04, + 1.0114e-04, 9.3205e-05], device='cuda:1') +2023-02-06 05:41:13,771 INFO [train.py:901] (1/4) Epoch 8, batch 3800, loss[loss=0.3307, simple_loss=0.3769, pruned_loss=0.1422, over 8546.00 frames. ], tot_loss[loss=0.2534, simple_loss=0.3241, pruned_loss=0.09132, over 1609598.05 frames. ], batch size: 49, lr: 9.61e-03, grad_scale: 8.0 +2023-02-06 05:41:28,150 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60402.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:41:36,471 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60414.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:41:45,874 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60427.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:41:49,063 INFO [train.py:901] (1/4) Epoch 8, batch 3850, loss[loss=0.2841, simple_loss=0.3631, pruned_loss=0.1025, over 8297.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3254, pruned_loss=0.09189, over 1616261.70 frames. ], batch size: 23, lr: 9.60e-03, grad_scale: 8.0 +2023-02-06 05:41:52,292 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.691e+02 3.271e+02 4.212e+02 1.032e+03, threshold=6.541e+02, percent-clipped=5.0 +2023-02-06 05:41:54,201 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60440.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:42:08,493 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 05:42:22,366 INFO [train.py:901] (1/4) Epoch 8, batch 3900, loss[loss=0.2802, simple_loss=0.353, pruned_loss=0.1037, over 8560.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3256, pruned_loss=0.09272, over 1615299.25 frames. ], batch size: 31, lr: 9.60e-03, grad_scale: 8.0 +2023-02-06 05:42:46,597 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60517.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:42:53,421 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6661, 1.7520, 1.9871, 1.6442, 1.0833, 2.0708, 0.1930, 1.2460], + device='cuda:1'), covar=tensor([0.3185, 0.2036, 0.0622, 0.2043, 0.6050, 0.0618, 0.4869, 0.2543], + device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0147, 0.0085, 0.0196, 0.0238, 0.0091, 0.0155, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:42:55,286 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60529.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:42:57,130 INFO [train.py:901] (1/4) Epoch 8, batch 3950, loss[loss=0.2566, simple_loss=0.3415, pruned_loss=0.08587, over 8260.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3243, pruned_loss=0.09157, over 1611848.86 frames. ], batch size: 24, lr: 9.59e-03, grad_scale: 8.0 +2023-02-06 05:43:00,412 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.562e+02 3.362e+02 4.082e+02 8.516e+02, threshold=6.724e+02, percent-clipped=2.0 +2023-02-06 05:43:03,868 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60542.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:43:13,333 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60555.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:43:16,391 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60560.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:43:28,571 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3569, 1.7319, 1.5756, 1.4178, 1.2144, 1.4651, 1.8722, 1.5745], + device='cuda:1'), covar=tensor([0.0442, 0.1098, 0.1591, 0.1251, 0.0531, 0.1377, 0.0626, 0.0606], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0161, 0.0198, 0.0163, 0.0111, 0.0169, 0.0123, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 05:43:31,325 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60581.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:43:31,761 INFO [train.py:901] (1/4) Epoch 8, batch 4000, loss[loss=0.2587, simple_loss=0.3353, pruned_loss=0.09109, over 8476.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3247, pruned_loss=0.09173, over 1610609.35 frames. ], batch size: 28, lr: 9.59e-03, grad_scale: 16.0 +2023-02-06 05:43:48,271 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60606.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:43:59,774 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:44:06,267 INFO [train.py:901] (1/4) Epoch 8, batch 4050, loss[loss=0.2596, simple_loss=0.3256, pruned_loss=0.09675, over 8012.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3248, pruned_loss=0.09148, over 1614015.40 frames. ], batch size: 22, lr: 9.59e-03, grad_scale: 16.0 +2023-02-06 05:44:09,642 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.835e+02 3.722e+02 4.462e+02 8.493e+02, threshold=7.445e+02, percent-clipped=1.0 +2023-02-06 05:44:17,202 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60647.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:44:24,713 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8763, 1.9480, 2.1766, 1.7696, 1.3231, 2.1428, 0.6189, 1.4988], + device='cuda:1'), covar=tensor([0.2491, 0.1749, 0.0668, 0.2083, 0.4516, 0.0613, 0.3957, 0.2181], + device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0145, 0.0085, 0.0193, 0.0232, 0.0090, 0.0151, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:44:30,905 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8062, 1.5070, 2.0328, 1.7480, 1.8101, 1.7158, 1.3664, 0.7184], + device='cuda:1'), covar=tensor([0.2548, 0.2546, 0.0782, 0.1462, 0.1206, 0.1334, 0.1207, 0.2447], + device='cuda:1'), in_proj_covar=tensor([0.0829, 0.0802, 0.0681, 0.0780, 0.0879, 0.0734, 0.0675, 0.0718], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:44:36,979 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60675.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:44:41,540 INFO [train.py:901] (1/4) Epoch 8, batch 4100, loss[loss=0.2442, simple_loss=0.3048, pruned_loss=0.09178, over 7440.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3243, pruned_loss=0.09151, over 1608657.73 frames. ], batch size: 17, lr: 9.58e-03, grad_scale: 16.0 +2023-02-06 05:45:16,312 INFO [train.py:901] (1/4) Epoch 8, batch 4150, loss[loss=0.3475, simple_loss=0.3871, pruned_loss=0.1539, over 7346.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3256, pruned_loss=0.09255, over 1609581.11 frames. ], batch size: 72, lr: 9.58e-03, grad_scale: 16.0 +2023-02-06 05:45:19,089 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:45:19,618 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.904e+02 3.574e+02 4.093e+02 8.234e+02, threshold=7.147e+02, percent-clipped=2.0 +2023-02-06 05:45:45,332 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60773.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:45:51,066 INFO [train.py:901] (1/4) Epoch 8, batch 4200, loss[loss=0.2833, simple_loss=0.348, pruned_loss=0.1093, over 8481.00 frames. ], tot_loss[loss=0.2567, simple_loss=0.3268, pruned_loss=0.09335, over 1606990.12 frames. ], batch size: 27, lr: 9.57e-03, grad_scale: 8.0 +2023-02-06 05:45:54,024 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:02,586 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60798.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:02,606 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60798.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:07,981 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 05:46:10,874 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:11,535 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60811.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:20,183 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60823.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:25,969 INFO [train.py:901] (1/4) Epoch 8, batch 4250, loss[loss=0.3508, simple_loss=0.3995, pruned_loss=0.151, over 6713.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3284, pruned_loss=0.09436, over 1606584.35 frames. ], batch size: 72, lr: 9.57e-03, grad_scale: 8.0 +2023-02-06 05:46:26,076 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6226, 5.8591, 4.8781, 2.0167, 5.0778, 5.4535, 5.4003, 4.8648], + device='cuda:1'), covar=tensor([0.0703, 0.0478, 0.1034, 0.5300, 0.0670, 0.0674, 0.1118, 0.0759], + device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0331, 0.0354, 0.0453, 0.0350, 0.0328, 0.0338, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:46:28,893 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60836.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:30,101 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.898e+02 2.807e+02 3.546e+02 4.515e+02 1.213e+03, threshold=7.092e+02, percent-clipped=3.0 +2023-02-06 05:46:30,847 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 05:47:01,077 INFO [train.py:901] (1/4) Epoch 8, batch 4300, loss[loss=0.2732, simple_loss=0.3547, pruned_loss=0.09592, over 8104.00 frames. ], tot_loss[loss=0.2576, simple_loss=0.3277, pruned_loss=0.09373, over 1611278.36 frames. ], batch size: 23, lr: 9.57e-03, grad_scale: 8.0 +2023-02-06 05:47:35,210 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60931.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:47:35,682 INFO [train.py:901] (1/4) Epoch 8, batch 4350, loss[loss=0.2615, simple_loss=0.3242, pruned_loss=0.09937, over 7791.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.327, pruned_loss=0.09361, over 1609129.32 frames. ], batch size: 19, lr: 9.56e-03, grad_scale: 8.0 +2023-02-06 05:47:39,637 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.751e+02 3.442e+02 4.335e+02 7.709e+02, threshold=6.884e+02, percent-clipped=1.0 +2023-02-06 05:47:51,973 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60956.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:47:59,293 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 05:48:07,094 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4480, 2.6506, 1.7070, 2.0176, 2.0758, 1.4251, 1.7733, 1.9955], + device='cuda:1'), covar=tensor([0.1241, 0.0278, 0.0902, 0.0558, 0.0636, 0.1172, 0.0941, 0.0873], + device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0231, 0.0306, 0.0290, 0.0300, 0.0308, 0.0333, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 05:48:10,332 INFO [train.py:901] (1/4) Epoch 8, batch 4400, loss[loss=0.2057, simple_loss=0.2873, pruned_loss=0.06203, over 8235.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3249, pruned_loss=0.0919, over 1612063.82 frames. ], batch size: 22, lr: 9.56e-03, grad_scale: 8.0 +2023-02-06 05:48:42,561 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 05:48:45,134 INFO [train.py:901] (1/4) Epoch 8, batch 4450, loss[loss=0.2297, simple_loss=0.3182, pruned_loss=0.07059, over 8486.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3235, pruned_loss=0.09098, over 1609755.85 frames. ], batch size: 48, lr: 9.55e-03, grad_scale: 8.0 +2023-02-06 05:48:49,126 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.403e+02 3.130e+02 3.948e+02 8.767e+02, threshold=6.260e+02, percent-clipped=3.0 +2023-02-06 05:49:17,370 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61079.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:49:17,969 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=61080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:49:19,304 INFO [train.py:901] (1/4) Epoch 8, batch 4500, loss[loss=0.2324, simple_loss=0.312, pruned_loss=0.07643, over 8464.00 frames. ], tot_loss[loss=0.2524, simple_loss=0.3228, pruned_loss=0.09097, over 1609361.02 frames. ], batch size: 29, lr: 9.55e-03, grad_scale: 8.0 +2023-02-06 05:49:36,534 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 05:49:53,484 INFO [train.py:901] (1/4) Epoch 8, batch 4550, loss[loss=0.2538, simple_loss=0.3243, pruned_loss=0.09171, over 7436.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3241, pruned_loss=0.09233, over 1611510.71 frames. ], batch size: 17, lr: 9.55e-03, grad_scale: 8.0 +2023-02-06 05:49:58,150 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.768e+02 3.493e+02 4.645e+02 1.007e+03, threshold=6.986e+02, percent-clipped=6.0 +2023-02-06 05:50:29,333 INFO [train.py:901] (1/4) Epoch 8, batch 4600, loss[loss=0.2142, simple_loss=0.2963, pruned_loss=0.06608, over 7928.00 frames. ], tot_loss[loss=0.253, simple_loss=0.3232, pruned_loss=0.09139, over 1611009.87 frames. ], batch size: 20, lr: 9.54e-03, grad_scale: 8.0 +2023-02-06 05:50:38,244 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61195.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:51:03,415 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.30 vs. limit=5.0 +2023-02-06 05:51:04,234 INFO [train.py:901] (1/4) Epoch 8, batch 4650, loss[loss=0.2426, simple_loss=0.3186, pruned_loss=0.08329, over 7912.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3233, pruned_loss=0.09142, over 1610032.08 frames. ], batch size: 20, lr: 9.54e-03, grad_scale: 8.0 +2023-02-06 05:51:08,278 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 2.693e+02 3.115e+02 3.876e+02 8.832e+02, threshold=6.229e+02, percent-clipped=3.0 +2023-02-06 05:51:13,729 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1483, 1.6770, 1.2423, 1.7735, 1.4116, 1.0138, 1.3016, 1.6045], + device='cuda:1'), covar=tensor([0.0832, 0.0396, 0.1129, 0.0401, 0.0585, 0.1341, 0.0679, 0.0593], + device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0234, 0.0311, 0.0297, 0.0305, 0.0316, 0.0340, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 05:51:38,690 INFO [train.py:901] (1/4) Epoch 8, batch 4700, loss[loss=0.2747, simple_loss=0.3464, pruned_loss=0.1015, over 8192.00 frames. ], tot_loss[loss=0.2524, simple_loss=0.3229, pruned_loss=0.09095, over 1610505.07 frames. ], batch size: 23, lr: 9.54e-03, grad_scale: 8.0 +2023-02-06 05:52:13,892 INFO [train.py:901] (1/4) Epoch 8, batch 4750, loss[loss=0.2256, simple_loss=0.3162, pruned_loss=0.06748, over 8354.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3247, pruned_loss=0.09135, over 1613031.79 frames. ], batch size: 24, lr: 9.53e-03, grad_scale: 8.0 +2023-02-06 05:52:17,856 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.870e+02 3.425e+02 4.672e+02 9.837e+02, threshold=6.850e+02, percent-clipped=8.0 +2023-02-06 05:52:36,777 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 05:52:39,424 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 05:52:48,203 INFO [train.py:901] (1/4) Epoch 8, batch 4800, loss[loss=0.2212, simple_loss=0.294, pruned_loss=0.07421, over 8765.00 frames. ], tot_loss[loss=0.254, simple_loss=0.3246, pruned_loss=0.09172, over 1613957.60 frames. ], batch size: 30, lr: 9.53e-03, grad_scale: 8.0 +2023-02-06 05:53:16,793 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=61423.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:53:22,688 INFO [train.py:901] (1/4) Epoch 8, batch 4850, loss[loss=0.2466, simple_loss=0.3293, pruned_loss=0.08197, over 8295.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.3231, pruned_loss=0.09059, over 1612955.42 frames. ], batch size: 23, lr: 9.52e-03, grad_scale: 8.0 +2023-02-06 05:53:26,648 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.608e+02 2.780e+02 3.448e+02 4.323e+02 7.771e+02, threshold=6.895e+02, percent-clipped=1.0 +2023-02-06 05:53:28,677 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 05:53:36,187 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61451.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:53:53,270 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61476.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:53:56,863 INFO [train.py:901] (1/4) Epoch 8, batch 4900, loss[loss=0.2248, simple_loss=0.2951, pruned_loss=0.07727, over 7533.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3247, pruned_loss=0.0916, over 1614826.86 frames. ], batch size: 18, lr: 9.52e-03, grad_scale: 8.0 +2023-02-06 05:54:31,293 INFO [train.py:901] (1/4) Epoch 8, batch 4950, loss[loss=0.2379, simple_loss=0.3111, pruned_loss=0.08241, over 8039.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.3252, pruned_loss=0.09216, over 1612759.70 frames. ], batch size: 22, lr: 9.52e-03, grad_scale: 8.0 +2023-02-06 05:54:35,319 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.817e+02 2.701e+02 3.325e+02 4.582e+02 7.633e+02, threshold=6.649e+02, percent-clipped=1.0 +2023-02-06 05:54:35,525 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61538.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:55:07,170 INFO [train.py:901] (1/4) Epoch 8, batch 5000, loss[loss=0.2487, simple_loss=0.3242, pruned_loss=0.08654, over 8322.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3254, pruned_loss=0.09241, over 1610538.71 frames. ], batch size: 26, lr: 9.51e-03, grad_scale: 8.0 +2023-02-06 05:55:20,035 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 05:55:42,157 INFO [train.py:901] (1/4) Epoch 8, batch 5050, loss[loss=0.2541, simple_loss=0.3356, pruned_loss=0.08629, over 8550.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3263, pruned_loss=0.09274, over 1611761.02 frames. ], batch size: 31, lr: 9.51e-03, grad_scale: 8.0 +2023-02-06 05:55:46,799 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.732e+02 2.771e+02 3.459e+02 4.924e+02 1.310e+03, threshold=6.919e+02, percent-clipped=9.0 +2023-02-06 05:56:07,690 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 05:56:17,395 INFO [train.py:901] (1/4) Epoch 8, batch 5100, loss[loss=0.2112, simple_loss=0.2937, pruned_loss=0.06437, over 8666.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3261, pruned_loss=0.09278, over 1613954.62 frames. ], batch size: 39, lr: 9.50e-03, grad_scale: 8.0 +2023-02-06 05:56:52,939 INFO [train.py:901] (1/4) Epoch 8, batch 5150, loss[loss=0.2673, simple_loss=0.3392, pruned_loss=0.09771, over 8531.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.326, pruned_loss=0.09277, over 1614434.45 frames. ], batch size: 34, lr: 9.50e-03, grad_scale: 8.0 +2023-02-06 05:56:57,120 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.584e+02 3.190e+02 4.018e+02 8.337e+02, threshold=6.381e+02, percent-clipped=2.0 +2023-02-06 05:57:02,454 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.19 vs. limit=5.0 +2023-02-06 05:57:06,181 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:57:13,688 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:57:27,556 INFO [train.py:901] (1/4) Epoch 8, batch 5200, loss[loss=0.2215, simple_loss=0.3064, pruned_loss=0.06824, over 8285.00 frames. ], tot_loss[loss=0.254, simple_loss=0.3243, pruned_loss=0.09184, over 1606616.94 frames. ], batch size: 23, lr: 9.50e-03, grad_scale: 8.0 +2023-02-06 05:57:35,700 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:57:53,056 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61819.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:58:02,400 INFO [train.py:901] (1/4) Epoch 8, batch 5250, loss[loss=0.2918, simple_loss=0.3641, pruned_loss=0.1097, over 8458.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.3247, pruned_loss=0.09183, over 1608742.92 frames. ], batch size: 27, lr: 9.49e-03, grad_scale: 8.0 +2023-02-06 05:58:06,458 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.578e+02 2.711e+02 3.309e+02 4.013e+02 1.150e+03, threshold=6.618e+02, percent-clipped=3.0 +2023-02-06 05:58:07,796 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 05:58:37,169 INFO [train.py:901] (1/4) Epoch 8, batch 5300, loss[loss=0.2345, simple_loss=0.3155, pruned_loss=0.07676, over 8456.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3254, pruned_loss=0.09269, over 1610720.31 frames. ], batch size: 27, lr: 9.49e-03, grad_scale: 8.0 +2023-02-06 05:59:04,765 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2894, 2.2698, 1.9613, 2.8036, 1.3048, 1.6317, 1.8184, 2.2483], + device='cuda:1'), covar=tensor([0.0670, 0.0830, 0.1074, 0.0372, 0.1352, 0.1567, 0.1194, 0.0832], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0226, 0.0267, 0.0216, 0.0228, 0.0260, 0.0264, 0.0234], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 05:59:12,105 INFO [train.py:901] (1/4) Epoch 8, batch 5350, loss[loss=0.2601, simple_loss=0.337, pruned_loss=0.0916, over 8372.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.324, pruned_loss=0.0919, over 1608327.58 frames. ], batch size: 24, lr: 9.49e-03, grad_scale: 8.0 +2023-02-06 05:59:12,262 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61932.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:59:16,961 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.596e+02 3.249e+02 3.983e+02 1.109e+03, threshold=6.498e+02, percent-clipped=6.0 +2023-02-06 05:59:17,996 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0036, 1.6254, 2.2991, 1.8987, 1.9586, 1.8911, 1.5102, 0.6763], + device='cuda:1'), covar=tensor([0.2886, 0.2862, 0.0822, 0.1714, 0.1352, 0.1584, 0.1286, 0.2802], + device='cuda:1'), in_proj_covar=tensor([0.0846, 0.0802, 0.0677, 0.0789, 0.0889, 0.0742, 0.0670, 0.0723], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 05:59:47,814 INFO [train.py:901] (1/4) Epoch 8, batch 5400, loss[loss=0.2158, simple_loss=0.2954, pruned_loss=0.0681, over 8084.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3247, pruned_loss=0.09252, over 1607421.11 frames. ], batch size: 21, lr: 9.48e-03, grad_scale: 8.0 +2023-02-06 06:00:23,964 INFO [train.py:901] (1/4) Epoch 8, batch 5450, loss[loss=0.2658, simple_loss=0.3462, pruned_loss=0.09266, over 8512.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3246, pruned_loss=0.09209, over 1609688.13 frames. ], batch size: 28, lr: 9.48e-03, grad_scale: 8.0 +2023-02-06 06:00:28,683 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.625e+02 3.240e+02 4.068e+02 8.471e+02, threshold=6.479e+02, percent-clipped=5.0 +2023-02-06 06:01:00,410 INFO [train.py:901] (1/4) Epoch 8, batch 5500, loss[loss=0.2359, simple_loss=0.308, pruned_loss=0.08196, over 8291.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3244, pruned_loss=0.09164, over 1610749.10 frames. ], batch size: 23, lr: 9.47e-03, grad_scale: 8.0 +2023-02-06 06:01:01,792 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 06:01:07,838 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3387, 1.5239, 4.4588, 1.6997, 3.8624, 3.7277, 4.0288, 3.9079], + device='cuda:1'), covar=tensor([0.0480, 0.3672, 0.0393, 0.2892, 0.1015, 0.0736, 0.0552, 0.0539], + device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0532, 0.0517, 0.0489, 0.0555, 0.0467, 0.0471, 0.0519], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 06:01:08,451 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:01:16,548 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62105.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:01:34,824 INFO [train.py:901] (1/4) Epoch 8, batch 5550, loss[loss=0.2358, simple_loss=0.3253, pruned_loss=0.07319, over 8351.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3244, pruned_loss=0.09169, over 1606612.38 frames. ], batch size: 25, lr: 9.47e-03, grad_scale: 8.0 +2023-02-06 06:01:38,602 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.724e+02 3.277e+02 4.222e+02 9.983e+02, threshold=6.553e+02, percent-clipped=5.0 +2023-02-06 06:01:40,086 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:02:09,237 INFO [train.py:901] (1/4) Epoch 8, batch 5600, loss[loss=0.2881, simple_loss=0.3509, pruned_loss=0.1126, over 5938.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.3249, pruned_loss=0.09232, over 1606028.79 frames. ], batch size: 13, lr: 9.47e-03, grad_scale: 8.0 +2023-02-06 06:02:27,945 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62209.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:02:35,390 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62220.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:02:44,121 INFO [train.py:901] (1/4) Epoch 8, batch 5650, loss[loss=0.2768, simple_loss=0.3619, pruned_loss=0.09584, over 8325.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3259, pruned_loss=0.09287, over 1607293.75 frames. ], batch size: 25, lr: 9.46e-03, grad_scale: 8.0 +2023-02-06 06:02:48,208 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.893e+02 3.442e+02 4.058e+02 7.819e+02, threshold=6.884e+02, percent-clipped=2.0 +2023-02-06 06:03:01,734 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 06:03:03,308 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 06:03:09,918 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 06:03:14,935 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62276.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:03:19,759 INFO [train.py:901] (1/4) Epoch 8, batch 5700, loss[loss=0.2309, simple_loss=0.3153, pruned_loss=0.07332, over 8552.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3247, pruned_loss=0.09236, over 1606734.67 frames. ], batch size: 31, lr: 9.46e-03, grad_scale: 8.0 +2023-02-06 06:03:24,177 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.57 vs. limit=5.0 +2023-02-06 06:03:26,022 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:03:44,792 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4120, 1.9084, 3.3347, 1.2261, 2.4562, 1.9161, 1.6055, 2.1947], + device='cuda:1'), covar=tensor([0.1681, 0.2041, 0.0765, 0.3639, 0.1528, 0.2577, 0.1653, 0.2208], + device='cuda:1'), in_proj_covar=tensor([0.0475, 0.0487, 0.0532, 0.0563, 0.0605, 0.0537, 0.0456, 0.0597], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:03:53,764 INFO [train.py:901] (1/4) Epoch 8, batch 5750, loss[loss=0.2565, simple_loss=0.3125, pruned_loss=0.1002, over 8137.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3244, pruned_loss=0.09204, over 1605281.74 frames. ], batch size: 22, lr: 9.45e-03, grad_scale: 8.0 +2023-02-06 06:03:58,445 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.702e+02 3.342e+02 4.214e+02 1.406e+03, threshold=6.684e+02, percent-clipped=3.0 +2023-02-06 06:04:07,822 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 06:04:28,274 INFO [train.py:901] (1/4) Epoch 8, batch 5800, loss[loss=0.2993, simple_loss=0.3568, pruned_loss=0.1209, over 7035.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3243, pruned_loss=0.09171, over 1610907.57 frames. ], batch size: 71, lr: 9.45e-03, grad_scale: 8.0 +2023-02-06 06:04:35,226 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:04:48,166 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62409.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:04,094 INFO [train.py:901] (1/4) Epoch 8, batch 5850, loss[loss=0.2543, simple_loss=0.3243, pruned_loss=0.09216, over 8609.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3238, pruned_loss=0.09138, over 1611011.84 frames. ], batch size: 34, lr: 9.45e-03, grad_scale: 8.0 +2023-02-06 06:05:08,211 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.970e+02 2.690e+02 3.286e+02 4.000e+02 6.740e+02, threshold=6.571e+02, percent-clipped=1.0 +2023-02-06 06:05:15,060 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2164, 1.1874, 1.2189, 1.1443, 0.8185, 1.3317, 0.0627, 1.0364], + device='cuda:1'), covar=tensor([0.2538, 0.1902, 0.0630, 0.1428, 0.4845, 0.0643, 0.3831, 0.1589], + device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0152, 0.0087, 0.0199, 0.0238, 0.0092, 0.0159, 0.0152], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 06:05:27,198 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62465.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:34,611 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62476.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:38,455 INFO [train.py:901] (1/4) Epoch 8, batch 5900, loss[loss=0.2976, simple_loss=0.3644, pruned_loss=0.1154, over 8457.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3239, pruned_loss=0.09158, over 1612128.38 frames. ], batch size: 29, lr: 9.44e-03, grad_scale: 8.0 +2023-02-06 06:05:39,872 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62484.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:44,017 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:51,361 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62501.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:54,884 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.77 vs. limit=5.0 +2023-02-06 06:06:06,966 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-02-06 06:06:13,413 INFO [train.py:901] (1/4) Epoch 8, batch 5950, loss[loss=0.1917, simple_loss=0.2697, pruned_loss=0.05688, over 7435.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3229, pruned_loss=0.09086, over 1610468.68 frames. ], batch size: 17, lr: 9.44e-03, grad_scale: 8.0 +2023-02-06 06:06:17,413 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 2.652e+02 3.220e+02 3.904e+02 8.315e+02, threshold=6.439e+02, percent-clipped=2.0 +2023-02-06 06:06:31,784 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6585, 2.0242, 3.6667, 1.3872, 2.7100, 2.1688, 1.7583, 2.3386], + device='cuda:1'), covar=tensor([0.1530, 0.2002, 0.0602, 0.3534, 0.1244, 0.2320, 0.1534, 0.2033], + device='cuda:1'), in_proj_covar=tensor([0.0479, 0.0490, 0.0535, 0.0568, 0.0604, 0.0540, 0.0460, 0.0603], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:06:34,795 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62563.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:06:47,824 INFO [train.py:901] (1/4) Epoch 8, batch 6000, loss[loss=0.2573, simple_loss=0.3333, pruned_loss=0.09071, over 8195.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3226, pruned_loss=0.09103, over 1608899.97 frames. ], batch size: 23, lr: 9.44e-03, grad_scale: 8.0 +2023-02-06 06:06:47,824 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 06:07:00,014 INFO [train.py:935] (1/4) Epoch 8, validation: loss=0.1996, simple_loss=0.2985, pruned_loss=0.05037, over 944034.00 frames. +2023-02-06 06:07:00,014 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 06:07:12,287 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62599.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:07:14,963 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:07:33,818 INFO [train.py:901] (1/4) Epoch 8, batch 6050, loss[loss=0.2441, simple_loss=0.3187, pruned_loss=0.08472, over 8454.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3216, pruned_loss=0.09034, over 1612198.81 frames. ], batch size: 25, lr: 9.43e-03, grad_scale: 8.0 +2023-02-06 06:07:35,913 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:07:35,987 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1691, 1.3153, 4.2450, 1.5743, 3.8490, 3.4910, 3.7992, 3.6672], + device='cuda:1'), covar=tensor([0.0424, 0.3579, 0.0437, 0.2788, 0.0861, 0.0733, 0.0486, 0.0585], + device='cuda:1'), in_proj_covar=tensor([0.0438, 0.0549, 0.0530, 0.0500, 0.0563, 0.0475, 0.0480, 0.0534], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 06:07:37,858 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.575e+02 3.268e+02 4.071e+02 9.720e+02, threshold=6.536e+02, percent-clipped=3.0 +2023-02-06 06:07:39,343 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3026, 1.6708, 3.3079, 1.4525, 2.2190, 3.6475, 3.6053, 3.1570], + device='cuda:1'), covar=tensor([0.0857, 0.1440, 0.0385, 0.1896, 0.0952, 0.0290, 0.0489, 0.0623], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0286, 0.0246, 0.0274, 0.0257, 0.0227, 0.0295, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 06:07:44,177 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62647.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:08:02,398 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:08:08,762 INFO [train.py:901] (1/4) Epoch 8, batch 6100, loss[loss=0.2326, simple_loss=0.3106, pruned_loss=0.07735, over 8513.00 frames. ], tot_loss[loss=0.2525, simple_loss=0.3233, pruned_loss=0.09084, over 1616206.29 frames. ], batch size: 26, lr: 9.43e-03, grad_scale: 8.0 +2023-02-06 06:08:10,158 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62684.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:08:12,272 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6281, 2.1638, 3.6719, 2.7322, 2.9282, 2.2501, 1.7550, 1.6063], + device='cuda:1'), covar=tensor([0.3007, 0.3728, 0.0843, 0.2283, 0.1951, 0.1892, 0.1516, 0.3965], + device='cuda:1'), in_proj_covar=tensor([0.0848, 0.0800, 0.0681, 0.0785, 0.0880, 0.0739, 0.0671, 0.0721], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:08:14,894 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7546, 1.8700, 2.0685, 1.7041, 1.1885, 2.1524, 0.3593, 1.3163], + device='cuda:1'), covar=tensor([0.3398, 0.1774, 0.0724, 0.2196, 0.5292, 0.0646, 0.4514, 0.2185], + device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0153, 0.0088, 0.0202, 0.0241, 0.0094, 0.0159, 0.0155], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 06:08:30,095 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0261, 1.3122, 1.3812, 1.1337, 1.0067, 1.3041, 1.6143, 1.8128], + device='cuda:1'), covar=tensor([0.0526, 0.1243, 0.1750, 0.1505, 0.0605, 0.1461, 0.0671, 0.0514], + device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0160, 0.0200, 0.0165, 0.0112, 0.0168, 0.0123, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 06:08:33,472 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3674, 1.7309, 1.6516, 0.7834, 1.7835, 1.3307, 0.3211, 1.5563], + device='cuda:1'), covar=tensor([0.0235, 0.0145, 0.0150, 0.0242, 0.0170, 0.0445, 0.0375, 0.0128], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0282, 0.0232, 0.0337, 0.0271, 0.0429, 0.0329, 0.0310], + device='cuda:1'), out_proj_covar=tensor([1.0717e-04, 8.3397e-05, 6.8682e-05, 1.0028e-04, 8.2210e-05, 1.3959e-04, + 9.9998e-05, 9.2975e-05], device='cuda:1') +2023-02-06 06:08:37,242 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 06:08:38,228 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-02-06 06:08:42,660 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1932, 1.7606, 2.7084, 2.1487, 2.4279, 1.9707, 1.5385, 0.8925], + device='cuda:1'), covar=tensor([0.3266, 0.3352, 0.0791, 0.1872, 0.1276, 0.1763, 0.1509, 0.3437], + device='cuda:1'), in_proj_covar=tensor([0.0857, 0.0806, 0.0684, 0.0786, 0.0885, 0.0742, 0.0675, 0.0725], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:08:43,085 INFO [train.py:901] (1/4) Epoch 8, batch 6150, loss[loss=0.2327, simple_loss=0.3115, pruned_loss=0.07692, over 8288.00 frames. ], tot_loss[loss=0.2525, simple_loss=0.3232, pruned_loss=0.09094, over 1615530.58 frames. ], batch size: 23, lr: 9.42e-03, grad_scale: 8.0 +2023-02-06 06:08:47,080 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.667e+02 3.544e+02 4.037e+02 8.376e+02, threshold=7.087e+02, percent-clipped=5.0 +2023-02-06 06:08:52,369 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8658, 1.2924, 2.9730, 1.2132, 2.0021, 3.1775, 3.3071, 2.5765], + device='cuda:1'), covar=tensor([0.1107, 0.1686, 0.0485, 0.2133, 0.1075, 0.0414, 0.0598, 0.0894], + device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0284, 0.0243, 0.0272, 0.0255, 0.0224, 0.0292, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 06:08:55,074 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:08:56,501 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3602, 1.6795, 2.7275, 1.1914, 1.9994, 1.7296, 1.4179, 1.7431], + device='cuda:1'), covar=tensor([0.1611, 0.1936, 0.0683, 0.3421, 0.1429, 0.2502, 0.1681, 0.1953], + device='cuda:1'), in_proj_covar=tensor([0.0478, 0.0491, 0.0535, 0.0564, 0.0603, 0.0539, 0.0458, 0.0602], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:08:57,036 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62753.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:09:17,709 INFO [train.py:901] (1/4) Epoch 8, batch 6200, loss[loss=0.2825, simple_loss=0.3541, pruned_loss=0.1055, over 8622.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3228, pruned_loss=0.09018, over 1617558.08 frames. ], batch size: 34, lr: 9.42e-03, grad_scale: 16.0 +2023-02-06 06:09:23,750 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62791.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:09:49,075 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.89 vs. limit=5.0 +2023-02-06 06:09:52,744 INFO [train.py:901] (1/4) Epoch 8, batch 6250, loss[loss=0.2416, simple_loss=0.3048, pruned_loss=0.08922, over 7660.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3231, pruned_loss=0.09054, over 1618942.53 frames. ], batch size: 19, lr: 9.42e-03, grad_scale: 16.0 +2023-02-06 06:09:56,741 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.905e+02 2.717e+02 3.222e+02 4.596e+02 9.217e+02, threshold=6.445e+02, percent-clipped=3.0 +2023-02-06 06:09:57,875 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 06:10:08,421 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62855.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:10:16,909 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:10:24,986 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62880.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:10:26,044 INFO [train.py:901] (1/4) Epoch 8, batch 6300, loss[loss=0.271, simple_loss=0.3488, pruned_loss=0.0966, over 8352.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3237, pruned_loss=0.09131, over 1617271.20 frames. ], batch size: 24, lr: 9.41e-03, grad_scale: 16.0 +2023-02-06 06:10:44,086 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62907.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:10:44,273 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1908, 1.7144, 2.5917, 2.0857, 2.3271, 1.9854, 1.5844, 0.9380], + device='cuda:1'), covar=tensor([0.2942, 0.3209, 0.0817, 0.1791, 0.1423, 0.1769, 0.1508, 0.3215], + device='cuda:1'), in_proj_covar=tensor([0.0862, 0.0815, 0.0688, 0.0797, 0.0898, 0.0749, 0.0678, 0.0734], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:10:58,980 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0114, 2.5045, 3.0302, 1.0750, 3.1989, 1.6390, 1.3268, 1.5993], + device='cuda:1'), covar=tensor([0.0487, 0.0240, 0.0174, 0.0434, 0.0202, 0.0479, 0.0537, 0.0338], + device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0286, 0.0234, 0.0341, 0.0274, 0.0434, 0.0335, 0.0315], + device='cuda:1'), out_proj_covar=tensor([1.0864e-04, 8.4565e-05, 6.8971e-05, 1.0117e-04, 8.2937e-05, 1.4138e-04, + 1.0202e-04, 9.4461e-05], device='cuda:1') +2023-02-06 06:11:01,296 INFO [train.py:901] (1/4) Epoch 8, batch 6350, loss[loss=0.2524, simple_loss=0.3113, pruned_loss=0.09678, over 7931.00 frames. ], tot_loss[loss=0.2524, simple_loss=0.3232, pruned_loss=0.09082, over 1618638.80 frames. ], batch size: 20, lr: 9.41e-03, grad_scale: 16.0 +2023-02-06 06:11:05,347 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.824e+02 3.504e+02 4.161e+02 7.437e+02, threshold=7.007e+02, percent-clipped=2.0 +2023-02-06 06:11:12,103 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62947.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:11:35,243 INFO [train.py:901] (1/4) Epoch 8, batch 6400, loss[loss=0.2445, simple_loss=0.3327, pruned_loss=0.0782, over 8252.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3225, pruned_loss=0.09022, over 1612307.99 frames. ], batch size: 24, lr: 9.41e-03, grad_scale: 16.0 +2023-02-06 06:11:52,121 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:03,492 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:07,444 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63028.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:09,582 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63031.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:10,063 INFO [train.py:901] (1/4) Epoch 8, batch 6450, loss[loss=0.1794, simple_loss=0.2539, pruned_loss=0.0525, over 7789.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3227, pruned_loss=0.0903, over 1610774.44 frames. ], batch size: 20, lr: 9.40e-03, grad_scale: 16.0 +2023-02-06 06:12:14,129 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.060e+02 2.983e+02 3.820e+02 5.218e+02 9.633e+02, threshold=7.640e+02, percent-clipped=4.0 +2023-02-06 06:12:26,417 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9192, 1.7145, 1.5761, 1.3294, 1.3531, 1.4843, 2.2838, 2.0072], + device='cuda:1'), covar=tensor([0.0512, 0.1242, 0.1808, 0.1437, 0.0615, 0.1489, 0.0628, 0.0546], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0160, 0.0198, 0.0166, 0.0112, 0.0168, 0.0122, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 06:12:31,369 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:45,222 INFO [train.py:901] (1/4) Epoch 8, batch 6500, loss[loss=0.2486, simple_loss=0.3233, pruned_loss=0.08695, over 7548.00 frames. ], tot_loss[loss=0.2525, simple_loss=0.3235, pruned_loss=0.09079, over 1610818.08 frames. ], batch size: 18, lr: 9.40e-03, grad_scale: 16.0 +2023-02-06 06:13:14,221 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63124.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:13:20,113 INFO [train.py:901] (1/4) Epoch 8, batch 6550, loss[loss=0.262, simple_loss=0.3358, pruned_loss=0.0941, over 8588.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3242, pruned_loss=0.09121, over 1613779.12 frames. ], batch size: 31, lr: 9.40e-03, grad_scale: 16.0 +2023-02-06 06:13:21,925 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 06:13:22,301 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:13:24,225 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.645e+02 3.116e+02 3.905e+02 9.747e+02, threshold=6.232e+02, percent-clipped=3.0 +2023-02-06 06:13:27,764 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63143.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:13:31,975 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63149.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:13:49,400 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 06:13:54,676 INFO [train.py:901] (1/4) Epoch 8, batch 6600, loss[loss=0.3403, simple_loss=0.3818, pruned_loss=0.1494, over 6426.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3246, pruned_loss=0.09151, over 1614089.46 frames. ], batch size: 71, lr: 9.39e-03, grad_scale: 16.0 +2023-02-06 06:14:07,399 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 06:14:29,751 INFO [train.py:901] (1/4) Epoch 8, batch 6650, loss[loss=0.2854, simple_loss=0.3353, pruned_loss=0.1178, over 8327.00 frames. ], tot_loss[loss=0.2534, simple_loss=0.3239, pruned_loss=0.09144, over 1614403.67 frames. ], batch size: 26, lr: 9.39e-03, grad_scale: 16.0 +2023-02-06 06:14:33,634 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 2.943e+02 3.528e+02 4.449e+02 1.178e+03, threshold=7.055e+02, percent-clipped=8.0 +2023-02-06 06:14:34,509 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63239.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:14:42,440 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:15:01,181 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:15:03,732 INFO [train.py:901] (1/4) Epoch 8, batch 6700, loss[loss=0.2604, simple_loss=0.3414, pruned_loss=0.08972, over 8193.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3243, pruned_loss=0.09108, over 1620183.84 frames. ], batch size: 23, lr: 9.38e-03, grad_scale: 16.0 +2023-02-06 06:15:19,273 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63303.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:15:29,353 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63318.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:15:38,600 INFO [train.py:901] (1/4) Epoch 8, batch 6750, loss[loss=0.2671, simple_loss=0.3375, pruned_loss=0.09836, over 8108.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3243, pruned_loss=0.09169, over 1612772.67 frames. ], batch size: 23, lr: 9.38e-03, grad_scale: 8.0 +2023-02-06 06:15:43,297 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.362e+02 2.603e+02 3.068e+02 3.707e+02 1.416e+03, threshold=6.136e+02, percent-clipped=3.0 +2023-02-06 06:15:46,218 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:16:13,027 INFO [train.py:901] (1/4) Epoch 8, batch 6800, loss[loss=0.2589, simple_loss=0.3367, pruned_loss=0.09054, over 8335.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3252, pruned_loss=0.09229, over 1615148.20 frames. ], batch size: 25, lr: 9.38e-03, grad_scale: 8.0 +2023-02-06 06:16:21,016 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 06:16:24,480 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63399.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:16:24,626 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.60 vs. limit=5.0 +2023-02-06 06:16:42,719 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63424.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:16:47,930 INFO [train.py:901] (1/4) Epoch 8, batch 6850, loss[loss=0.2014, simple_loss=0.2774, pruned_loss=0.06267, over 8030.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3244, pruned_loss=0.09208, over 1612490.93 frames. ], batch size: 22, lr: 9.37e-03, grad_scale: 8.0 +2023-02-06 06:16:52,294 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:16:52,779 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.669e+02 3.418e+02 4.059e+02 7.847e+02, threshold=6.836e+02, percent-clipped=4.0 +2023-02-06 06:16:53,660 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63440.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:17:11,112 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 06:17:22,262 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-06 06:17:22,530 INFO [train.py:901] (1/4) Epoch 8, batch 6900, loss[loss=0.2842, simple_loss=0.3556, pruned_loss=0.1064, over 7936.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3243, pruned_loss=0.09226, over 1608241.43 frames. ], batch size: 20, lr: 9.37e-03, grad_scale: 8.0 +2023-02-06 06:17:39,556 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63506.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:17:58,287 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:17:58,755 INFO [train.py:901] (1/4) Epoch 8, batch 6950, loss[loss=0.2161, simple_loss=0.2906, pruned_loss=0.07077, over 7639.00 frames. ], tot_loss[loss=0.254, simple_loss=0.3244, pruned_loss=0.09182, over 1612993.22 frames. ], batch size: 19, lr: 9.37e-03, grad_scale: 8.0 +2023-02-06 06:18:03,569 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.671e+02 3.369e+02 4.495e+02 9.890e+02, threshold=6.738e+02, percent-clipped=4.0 +2023-02-06 06:18:18,620 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 06:18:32,018 INFO [train.py:901] (1/4) Epoch 8, batch 7000, loss[loss=0.2403, simple_loss=0.3227, pruned_loss=0.07894, over 8575.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3238, pruned_loss=0.09135, over 1612838.14 frames. ], batch size: 39, lr: 9.36e-03, grad_scale: 8.0 +2023-02-06 06:18:32,816 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:18:36,333 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:18:48,879 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63604.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:19:08,082 INFO [train.py:901] (1/4) Epoch 8, batch 7050, loss[loss=0.2422, simple_loss=0.3263, pruned_loss=0.0791, over 8189.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3224, pruned_loss=0.09033, over 1607732.61 frames. ], batch size: 23, lr: 9.36e-03, grad_scale: 8.0 +2023-02-06 06:19:12,569 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.720e+02 3.242e+02 3.930e+02 7.648e+02, threshold=6.484e+02, percent-clipped=3.0 +2023-02-06 06:19:42,435 INFO [train.py:901] (1/4) Epoch 8, batch 7100, loss[loss=0.2825, simple_loss=0.3419, pruned_loss=0.1115, over 8457.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3227, pruned_loss=0.09066, over 1610173.19 frames. ], batch size: 25, lr: 9.35e-03, grad_scale: 8.0 +2023-02-06 06:19:53,538 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:20:17,190 INFO [train.py:901] (1/4) Epoch 8, batch 7150, loss[loss=0.2596, simple_loss=0.3138, pruned_loss=0.1027, over 8054.00 frames. ], tot_loss[loss=0.2517, simple_loss=0.3222, pruned_loss=0.09063, over 1608766.20 frames. ], batch size: 20, lr: 9.35e-03, grad_scale: 8.0 +2023-02-06 06:20:21,748 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.778e+02 3.549e+02 4.516e+02 1.097e+03, threshold=7.098e+02, percent-clipped=7.0 +2023-02-06 06:20:36,024 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 06:20:51,752 INFO [train.py:901] (1/4) Epoch 8, batch 7200, loss[loss=0.2383, simple_loss=0.316, pruned_loss=0.0803, over 8136.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3217, pruned_loss=0.09046, over 1608717.95 frames. ], batch size: 22, lr: 9.35e-03, grad_scale: 8.0 +2023-02-06 06:20:51,835 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63782.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:20:53,162 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63784.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:20:59,479 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2440, 1.4522, 1.5619, 1.3775, 1.0243, 1.6239, 1.5702, 1.6551], + device='cuda:1'), covar=tensor([0.0508, 0.1285, 0.1825, 0.1414, 0.0614, 0.1382, 0.0735, 0.0595], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0160, 0.0198, 0.0163, 0.0111, 0.0167, 0.0122, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 06:21:23,980 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63828.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:21:26,386 INFO [train.py:901] (1/4) Epoch 8, batch 7250, loss[loss=0.2862, simple_loss=0.3514, pruned_loss=0.1105, over 8100.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3222, pruned_loss=0.0909, over 1610492.05 frames. ], batch size: 23, lr: 9.34e-03, grad_scale: 8.0 +2023-02-06 06:21:30,981 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.694e+02 3.202e+02 4.148e+02 8.009e+02, threshold=6.403e+02, percent-clipped=2.0 +2023-02-06 06:21:47,991 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63863.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:00,780 INFO [train.py:901] (1/4) Epoch 8, batch 7300, loss[loss=0.1933, simple_loss=0.2682, pruned_loss=0.05918, over 7200.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3211, pruned_loss=0.09025, over 1604262.94 frames. ], batch size: 16, lr: 9.34e-03, grad_scale: 8.0 +2023-02-06 06:22:12,956 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63897.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:14,359 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63899.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:36,720 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63931.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:37,312 INFO [train.py:901] (1/4) Epoch 8, batch 7350, loss[loss=0.275, simple_loss=0.3515, pruned_loss=0.09921, over 8328.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3216, pruned_loss=0.09012, over 1606688.30 frames. ], batch size: 25, lr: 9.34e-03, grad_scale: 8.0 +2023-02-06 06:22:38,891 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63934.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:42,243 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.551e+02 3.183e+02 3.767e+02 5.416e+02, threshold=6.365e+02, percent-clipped=0.0 +2023-02-06 06:22:48,939 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63948.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:52,632 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-02-06 06:22:53,083 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5927, 4.6321, 4.1556, 1.7531, 4.0489, 3.9332, 4.2218, 3.7782], + device='cuda:1'), covar=tensor([0.0711, 0.0511, 0.1043, 0.4923, 0.0881, 0.0805, 0.1264, 0.0709], + device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0349, 0.0366, 0.0461, 0.0359, 0.0339, 0.0353, 0.0299], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:22:53,943 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63954.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:23:05,949 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 06:23:10,851 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:23:12,560 INFO [train.py:901] (1/4) Epoch 8, batch 7400, loss[loss=0.2687, simple_loss=0.3382, pruned_loss=0.09954, over 8023.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3225, pruned_loss=0.09073, over 1607594.45 frames. ], batch size: 22, lr: 9.33e-03, grad_scale: 8.0 +2023-02-06 06:23:24,825 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 06:23:48,438 INFO [train.py:901] (1/4) Epoch 8, batch 7450, loss[loss=0.2798, simple_loss=0.3438, pruned_loss=0.1079, over 8554.00 frames. ], tot_loss[loss=0.2525, simple_loss=0.3225, pruned_loss=0.09122, over 1608191.36 frames. ], batch size: 31, lr: 9.33e-03, grad_scale: 8.0 +2023-02-06 06:23:50,937 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 06:23:53,163 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.827e+02 3.358e+02 3.935e+02 9.777e+02, threshold=6.715e+02, percent-clipped=5.0 +2023-02-06 06:23:54,697 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64041.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:23:58,109 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64046.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:24:05,483 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 06:24:10,172 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64063.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:24:23,174 INFO [train.py:901] (1/4) Epoch 8, batch 7500, loss[loss=0.2514, simple_loss=0.3206, pruned_loss=0.09112, over 8192.00 frames. ], tot_loss[loss=0.2518, simple_loss=0.3224, pruned_loss=0.09057, over 1611609.57 frames. ], batch size: 23, lr: 9.33e-03, grad_scale: 8.0 +2023-02-06 06:24:30,452 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2670, 1.7072, 1.5838, 1.5927, 1.2901, 1.6423, 2.0970, 2.0426], + device='cuda:1'), covar=tensor([0.0456, 0.1241, 0.1864, 0.1357, 0.0659, 0.1588, 0.0694, 0.0593], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0160, 0.0199, 0.0164, 0.0111, 0.0168, 0.0122, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 06:24:55,913 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-06 06:24:57,464 INFO [train.py:901] (1/4) Epoch 8, batch 7550, loss[loss=0.2553, simple_loss=0.3372, pruned_loss=0.08671, over 8294.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3243, pruned_loss=0.09171, over 1614661.36 frames. ], batch size: 23, lr: 9.32e-03, grad_scale: 8.0 +2023-02-06 06:25:02,130 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.846e+02 3.017e+02 3.905e+02 4.969e+02 7.546e+02, threshold=7.810e+02, percent-clipped=1.0 +2023-02-06 06:25:11,681 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:13,076 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64155.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:24,182 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64172.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:28,963 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64178.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:30,994 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64180.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:32,164 INFO [train.py:901] (1/4) Epoch 8, batch 7600, loss[loss=0.2596, simple_loss=0.3303, pruned_loss=0.09446, over 8323.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3234, pruned_loss=0.091, over 1617138.09 frames. ], batch size: 25, lr: 9.32e-03, grad_scale: 8.0 +2023-02-06 06:25:49,061 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64207.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:26:07,009 INFO [train.py:901] (1/4) Epoch 8, batch 7650, loss[loss=0.2506, simple_loss=0.3259, pruned_loss=0.08772, over 8351.00 frames. ], tot_loss[loss=0.2529, simple_loss=0.3236, pruned_loss=0.09109, over 1613543.49 frames. ], batch size: 24, lr: 9.31e-03, grad_scale: 8.0 +2023-02-06 06:26:11,826 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.778e+02 3.467e+02 5.154e+02 1.113e+03, threshold=6.933e+02, percent-clipped=3.0 +2023-02-06 06:26:19,378 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:26:38,169 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:26:40,891 INFO [train.py:901] (1/4) Epoch 8, batch 7700, loss[loss=0.2555, simple_loss=0.3261, pruned_loss=0.09249, over 8469.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3221, pruned_loss=0.09058, over 1612038.07 frames. ], batch size: 27, lr: 9.31e-03, grad_scale: 8.0 +2023-02-06 06:26:45,250 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64287.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:26:56,495 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64302.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:08,131 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64319.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:10,004 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:10,460 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 06:27:13,351 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:16,533 INFO [train.py:901] (1/4) Epoch 8, batch 7750, loss[loss=0.2303, simple_loss=0.3048, pruned_loss=0.07787, over 7963.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3215, pruned_loss=0.09058, over 1607725.16 frames. ], batch size: 21, lr: 9.31e-03, grad_scale: 8.0 +2023-02-06 06:27:21,023 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.379e+02 2.516e+02 3.070e+02 3.996e+02 6.859e+02, threshold=6.139e+02, percent-clipped=0.0 +2023-02-06 06:27:25,266 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64344.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:38,600 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64363.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:43,238 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8964, 1.2493, 1.5300, 1.2651, 0.8917, 1.2995, 1.5107, 1.4708], + device='cuda:1'), covar=tensor([0.0505, 0.1284, 0.1633, 0.1385, 0.0600, 0.1501, 0.0656, 0.0579], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0160, 0.0198, 0.0165, 0.0111, 0.0168, 0.0121, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 06:27:51,130 INFO [train.py:901] (1/4) Epoch 8, batch 7800, loss[loss=0.2572, simple_loss=0.3313, pruned_loss=0.09152, over 8035.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3224, pruned_loss=0.09085, over 1605465.88 frames. ], batch size: 22, lr: 9.30e-03, grad_scale: 8.0 +2023-02-06 06:27:53,240 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64385.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:58,818 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64393.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:28:25,447 INFO [train.py:901] (1/4) Epoch 8, batch 7850, loss[loss=0.2975, simple_loss=0.357, pruned_loss=0.119, over 8256.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3235, pruned_loss=0.09081, over 1613928.95 frames. ], batch size: 24, lr: 9.30e-03, grad_scale: 8.0 +2023-02-06 06:28:27,099 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-02-06 06:28:30,098 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.075e+02 2.873e+02 3.519e+02 4.505e+02 1.254e+03, threshold=7.037e+02, percent-clipped=6.0 +2023-02-06 06:28:31,631 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7644, 2.0047, 1.9779, 1.5246, 2.0821, 1.5626, 1.1637, 1.7711], + device='cuda:1'), covar=tensor([0.0260, 0.0147, 0.0108, 0.0228, 0.0154, 0.0351, 0.0366, 0.0137], + device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0286, 0.0237, 0.0350, 0.0276, 0.0444, 0.0338, 0.0321], + device='cuda:1'), out_proj_covar=tensor([1.0975e-04, 8.3635e-05, 6.9924e-05, 1.0379e-04, 8.2563e-05, 1.4448e-04, + 1.0234e-04, 9.5831e-05], device='cuda:1') +2023-02-06 06:28:58,103 INFO [train.py:901] (1/4) Epoch 8, batch 7900, loss[loss=0.3247, simple_loss=0.3688, pruned_loss=0.1403, over 6536.00 frames. ], tot_loss[loss=0.2534, simple_loss=0.3237, pruned_loss=0.09158, over 1610567.22 frames. ], batch size: 71, lr: 9.30e-03, grad_scale: 8.0 +2023-02-06 06:28:58,917 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64483.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:29:10,202 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64500.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:29:32,352 INFO [train.py:901] (1/4) Epoch 8, batch 7950, loss[loss=0.2477, simple_loss=0.3284, pruned_loss=0.08355, over 8588.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3258, pruned_loss=0.09237, over 1616492.12 frames. ], batch size: 39, lr: 9.29e-03, grad_scale: 8.0 +2023-02-06 06:29:37,068 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.689e+02 3.383e+02 4.341e+02 8.251e+02, threshold=6.766e+02, percent-clipped=4.0 +2023-02-06 06:29:40,082 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:29:54,374 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.52 vs. limit=5.0 +2023-02-06 06:29:56,713 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64568.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:30:03,530 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64578.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:30:06,048 INFO [train.py:901] (1/4) Epoch 8, batch 8000, loss[loss=0.2283, simple_loss=0.3121, pruned_loss=0.07223, over 8468.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3254, pruned_loss=0.09219, over 1618944.06 frames. ], batch size: 29, lr: 9.29e-03, grad_scale: 8.0 +2023-02-06 06:30:14,266 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:30:17,199 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6029, 2.0933, 3.5236, 1.2493, 2.3903, 1.9739, 1.6985, 2.1014], + device='cuda:1'), covar=tensor([0.1499, 0.1762, 0.0535, 0.3378, 0.1339, 0.2359, 0.1516, 0.2041], + device='cuda:1'), in_proj_covar=tensor([0.0469, 0.0478, 0.0516, 0.0553, 0.0591, 0.0527, 0.0451, 0.0588], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:30:20,524 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:30:23,051 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:30:26,235 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9921, 1.7648, 3.4475, 1.4264, 2.4833, 3.8889, 3.9668, 3.3752], + device='cuda:1'), covar=tensor([0.1037, 0.1415, 0.0371, 0.1916, 0.0818, 0.0226, 0.0362, 0.0558], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0284, 0.0242, 0.0274, 0.0251, 0.0225, 0.0295, 0.0283], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 06:30:39,928 INFO [train.py:901] (1/4) Epoch 8, batch 8050, loss[loss=0.3024, simple_loss=0.357, pruned_loss=0.1239, over 6840.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3241, pruned_loss=0.09203, over 1606469.61 frames. ], batch size: 71, lr: 9.29e-03, grad_scale: 8.0 +2023-02-06 06:30:41,681 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 06:30:44,642 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 2.434e+02 2.955e+02 3.616e+02 6.730e+02, threshold=5.909e+02, percent-clipped=0.0 +2023-02-06 06:30:51,580 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:31:13,265 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 06:31:17,613 INFO [train.py:901] (1/4) Epoch 9, batch 0, loss[loss=0.255, simple_loss=0.3368, pruned_loss=0.08663, over 8462.00 frames. ], tot_loss[loss=0.255, simple_loss=0.3368, pruned_loss=0.08663, over 8462.00 frames. ], batch size: 25, lr: 8.79e-03, grad_scale: 8.0 +2023-02-06 06:31:17,614 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 06:31:28,851 INFO [train.py:935] (1/4) Epoch 9, validation: loss=0.1983, simple_loss=0.2974, pruned_loss=0.04961, over 944034.00 frames. +2023-02-06 06:31:28,852 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 06:31:29,659 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64666.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:31:35,196 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64674.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:31:35,994 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 +2023-02-06 06:31:43,423 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 06:31:56,868 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64707.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:31:58,420 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:32:02,438 INFO [train.py:901] (1/4) Epoch 9, batch 50, loss[loss=0.2125, simple_loss=0.2986, pruned_loss=0.06322, over 8472.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.3241, pruned_loss=0.0901, over 368025.82 frames. ], batch size: 25, lr: 8.79e-03, grad_scale: 8.0 +2023-02-06 06:32:06,627 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64721.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:32:07,944 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4535, 1.9880, 3.3585, 1.1441, 2.3137, 1.7517, 1.7421, 1.9378], + device='cuda:1'), covar=tensor([0.1966, 0.2100, 0.0754, 0.4194, 0.1732, 0.3105, 0.1823, 0.2545], + device='cuda:1'), in_proj_covar=tensor([0.0475, 0.0482, 0.0522, 0.0562, 0.0595, 0.0532, 0.0457, 0.0595], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:32:16,341 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 06:32:18,980 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.818e+02 3.347e+02 4.122e+02 1.189e+03, threshold=6.695e+02, percent-clipped=9.0 +2023-02-06 06:32:30,208 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4365, 1.8580, 1.8557, 1.0305, 1.9620, 1.3329, 0.5078, 1.6189], + device='cuda:1'), covar=tensor([0.0287, 0.0161, 0.0122, 0.0289, 0.0181, 0.0459, 0.0423, 0.0143], + device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0288, 0.0239, 0.0353, 0.0281, 0.0447, 0.0340, 0.0324], + device='cuda:1'), out_proj_covar=tensor([1.1223e-04, 8.4071e-05, 7.0473e-05, 1.0408e-04, 8.4099e-05, 1.4525e-04, + 1.0258e-04, 9.6760e-05], device='cuda:1') +2023-02-06 06:32:31,576 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64756.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:32:36,073 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64763.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:32:37,300 INFO [train.py:901] (1/4) Epoch 9, batch 100, loss[loss=0.241, simple_loss=0.3237, pruned_loss=0.07911, over 8493.00 frames. ], tot_loss[loss=0.2514, simple_loss=0.3227, pruned_loss=0.09003, over 641960.15 frames. ], batch size: 28, lr: 8.78e-03, grad_scale: 8.0 +2023-02-06 06:32:41,535 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64770.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:32:42,071 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 06:32:49,741 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64781.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:33:11,715 INFO [train.py:901] (1/4) Epoch 9, batch 150, loss[loss=0.3035, simple_loss=0.3481, pruned_loss=0.1295, over 7342.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3237, pruned_loss=0.09126, over 859657.98 frames. ], batch size: 71, lr: 8.78e-03, grad_scale: 8.0 +2023-02-06 06:33:16,753 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64822.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:33:20,058 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64827.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:33:27,775 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.577e+02 3.213e+02 3.848e+02 9.281e+02, threshold=6.425e+02, percent-clipped=3.0 +2023-02-06 06:33:45,631 INFO [train.py:901] (1/4) Epoch 9, batch 200, loss[loss=0.2878, simple_loss=0.3484, pruned_loss=0.1136, over 8367.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3247, pruned_loss=0.09139, over 1030523.14 frames. ], batch size: 24, lr: 8.78e-03, grad_scale: 8.0 +2023-02-06 06:34:21,132 INFO [train.py:901] (1/4) Epoch 9, batch 250, loss[loss=0.231, simple_loss=0.2963, pruned_loss=0.08283, over 7547.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.3245, pruned_loss=0.0913, over 1154124.43 frames. ], batch size: 18, lr: 8.77e-03, grad_scale: 8.0 +2023-02-06 06:34:34,285 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 06:34:36,819 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.096e+02 2.841e+02 3.295e+02 4.179e+02 1.029e+03, threshold=6.590e+02, percent-clipped=5.0 +2023-02-06 06:34:39,012 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64942.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:34:42,833 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 06:34:44,900 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:34:54,260 INFO [train.py:901] (1/4) Epoch 9, batch 300, loss[loss=0.2023, simple_loss=0.2954, pruned_loss=0.05456, over 8461.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3245, pruned_loss=0.0906, over 1259448.03 frames. ], batch size: 25, lr: 8.77e-03, grad_scale: 8.0 +2023-02-06 06:34:54,455 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:35:11,900 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64990.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:35:15,052 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64994.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:35:26,440 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65010.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:35:30,327 INFO [train.py:901] (1/4) Epoch 9, batch 350, loss[loss=0.2752, simple_loss=0.3447, pruned_loss=0.1029, over 8633.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3243, pruned_loss=0.09057, over 1337216.88 frames. ], batch size: 49, lr: 8.77e-03, grad_scale: 8.0 +2023-02-06 06:35:31,388 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 06:35:46,490 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.570e+02 3.183e+02 3.796e+02 1.000e+03, threshold=6.367e+02, percent-clipped=4.0 +2023-02-06 06:36:03,872 INFO [train.py:901] (1/4) Epoch 9, batch 400, loss[loss=0.1928, simple_loss=0.276, pruned_loss=0.05482, over 8027.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3219, pruned_loss=0.08881, over 1399334.74 frames. ], batch size: 22, lr: 8.76e-03, grad_scale: 8.0 +2023-02-06 06:36:03,940 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65065.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:04,747 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:09,398 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65073.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:12,871 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65078.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:30,518 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65103.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:33,062 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:37,774 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65114.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:38,389 INFO [train.py:901] (1/4) Epoch 9, batch 450, loss[loss=0.2452, simple_loss=0.3122, pruned_loss=0.08906, over 7644.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3219, pruned_loss=0.08879, over 1449241.48 frames. ], batch size: 19, lr: 8.76e-03, grad_scale: 8.0 +2023-02-06 06:36:46,200 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:51,803 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4382, 1.8564, 1.8529, 0.9880, 2.0268, 1.3885, 0.4042, 1.6585], + device='cuda:1'), covar=tensor([0.0285, 0.0161, 0.0140, 0.0271, 0.0194, 0.0518, 0.0440, 0.0132], + device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0291, 0.0240, 0.0349, 0.0281, 0.0445, 0.0338, 0.0321], + device='cuda:1'), out_proj_covar=tensor([1.1022e-04, 8.5274e-05, 7.0688e-05, 1.0267e-04, 8.4168e-05, 1.4412e-04, + 1.0199e-04, 9.5694e-05], device='cuda:1') +2023-02-06 06:36:53,165 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0736, 1.3058, 4.2155, 1.4515, 3.7025, 3.4086, 3.8133, 3.6152], + device='cuda:1'), covar=tensor([0.0566, 0.4481, 0.0490, 0.3371, 0.1125, 0.0859, 0.0561, 0.0724], + device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0548, 0.0538, 0.0496, 0.0562, 0.0476, 0.0479, 0.0535], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 06:36:56,313 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.705e+02 3.323e+02 3.920e+02 9.407e+02, threshold=6.647e+02, percent-clipped=6.0 +2023-02-06 06:37:08,207 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7694, 1.3783, 3.9024, 1.3458, 3.4562, 3.2078, 3.4985, 3.3457], + device='cuda:1'), covar=tensor([0.0539, 0.3780, 0.0593, 0.3186, 0.1133, 0.0856, 0.0592, 0.0692], + device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0547, 0.0535, 0.0494, 0.0560, 0.0476, 0.0477, 0.0532], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 06:37:13,479 INFO [train.py:901] (1/4) Epoch 9, batch 500, loss[loss=0.2286, simple_loss=0.3134, pruned_loss=0.07188, over 8191.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.3226, pruned_loss=0.08864, over 1486926.40 frames. ], batch size: 23, lr: 8.76e-03, grad_scale: 8.0 +2023-02-06 06:37:23,274 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65180.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:37:35,046 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65198.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:37:45,904 INFO [train.py:901] (1/4) Epoch 9, batch 550, loss[loss=0.2718, simple_loss=0.3441, pruned_loss=0.09975, over 8452.00 frames. ], tot_loss[loss=0.2519, simple_loss=0.3232, pruned_loss=0.0903, over 1510981.72 frames. ], batch size: 27, lr: 8.75e-03, grad_scale: 8.0 +2023-02-06 06:37:51,955 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65222.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:37:52,663 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65223.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:37:52,681 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0942, 1.2931, 1.2474, 0.7548, 1.3338, 1.0466, 0.4041, 1.1644], + device='cuda:1'), covar=tensor([0.0180, 0.0132, 0.0107, 0.0195, 0.0118, 0.0307, 0.0299, 0.0108], + device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0290, 0.0239, 0.0348, 0.0279, 0.0441, 0.0336, 0.0320], + device='cuda:1'), out_proj_covar=tensor([1.0949e-04, 8.5071e-05, 7.0429e-05, 1.0219e-04, 8.3463e-05, 1.4277e-04, + 1.0133e-04, 9.5413e-05], device='cuda:1') +2023-02-06 06:37:56,727 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65229.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:38:03,152 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.829e+02 3.496e+02 4.355e+02 8.306e+02, threshold=6.991e+02, percent-clipped=2.0 +2023-02-06 06:38:21,250 INFO [train.py:901] (1/4) Epoch 9, batch 600, loss[loss=0.2372, simple_loss=0.3062, pruned_loss=0.08415, over 7809.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.3221, pruned_loss=0.0896, over 1533085.63 frames. ], batch size: 20, lr: 8.75e-03, grad_scale: 8.0 +2023-02-06 06:38:38,352 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 06:38:43,860 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1166, 2.7219, 3.1092, 1.4244, 3.2682, 1.9589, 1.5391, 2.0359], + device='cuda:1'), covar=tensor([0.0486, 0.0180, 0.0192, 0.0411, 0.0227, 0.0483, 0.0546, 0.0275], + device='cuda:1'), in_proj_covar=tensor([0.0360, 0.0286, 0.0237, 0.0345, 0.0277, 0.0440, 0.0333, 0.0315], + device='cuda:1'), out_proj_covar=tensor([1.0834e-04, 8.3912e-05, 6.9865e-05, 1.0129e-04, 8.3041e-05, 1.4269e-04, + 1.0025e-04, 9.3990e-05], device='cuda:1') +2023-02-06 06:38:46,500 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65303.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:38:54,293 INFO [train.py:901] (1/4) Epoch 9, batch 650, loss[loss=0.2485, simple_loss=0.3279, pruned_loss=0.08452, over 8471.00 frames. ], tot_loss[loss=0.2524, simple_loss=0.3233, pruned_loss=0.0907, over 1545863.21 frames. ], batch size: 25, lr: 8.75e-03, grad_scale: 16.0 +2023-02-06 06:38:58,620 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 06:38:59,108 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:39:10,303 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:39:10,871 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.656e+02 3.252e+02 4.080e+02 6.220e+02, threshold=6.503e+02, percent-clipped=0.0 +2023-02-06 06:39:17,065 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:39:29,074 INFO [train.py:901] (1/4) Epoch 9, batch 700, loss[loss=0.2224, simple_loss=0.297, pruned_loss=0.07396, over 7665.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3212, pruned_loss=0.08908, over 1560465.75 frames. ], batch size: 19, lr: 8.74e-03, grad_scale: 16.0 +2023-02-06 06:39:41,003 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65381.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:39:57,977 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65406.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:03,854 INFO [train.py:901] (1/4) Epoch 9, batch 750, loss[loss=0.239, simple_loss=0.3225, pruned_loss=0.07779, over 8452.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3226, pruned_loss=0.09004, over 1569089.02 frames. ], batch size: 27, lr: 8.74e-03, grad_scale: 16.0 +2023-02-06 06:40:05,336 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:10,998 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-02-06 06:40:18,170 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65436.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:19,959 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.803e+02 3.527e+02 4.474e+02 1.505e+03, threshold=7.053e+02, percent-clipped=7.0 +2023-02-06 06:40:21,305 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 06:40:25,581 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.50 vs. limit=5.0 +2023-02-06 06:40:30,033 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 06:40:30,193 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65453.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:36,146 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65461.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:38,585 INFO [train.py:901] (1/4) Epoch 9, batch 800, loss[loss=0.1912, simple_loss=0.2703, pruned_loss=0.05602, over 7804.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.323, pruned_loss=0.08966, over 1583854.85 frames. ], batch size: 19, lr: 8.74e-03, grad_scale: 16.0 +2023-02-06 06:40:47,458 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:53,603 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65485.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:58,219 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1121, 2.3097, 1.8711, 2.9768, 1.3960, 1.5023, 2.0000, 2.4720], + device='cuda:1'), covar=tensor([0.0835, 0.0971, 0.1106, 0.0397, 0.1283, 0.1808, 0.1148, 0.0760], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0226, 0.0270, 0.0219, 0.0226, 0.0263, 0.0266, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 06:41:05,668 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65503.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:41:09,183 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3763, 1.9361, 3.1437, 2.3365, 2.6251, 2.1399, 1.6440, 1.3375], + device='cuda:1'), covar=tensor([0.3234, 0.3345, 0.0842, 0.2220, 0.1769, 0.1690, 0.1480, 0.3616], + device='cuda:1'), in_proj_covar=tensor([0.0855, 0.0810, 0.0700, 0.0803, 0.0895, 0.0751, 0.0685, 0.0729], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:41:10,492 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:41:13,591 INFO [train.py:901] (1/4) Epoch 9, batch 850, loss[loss=0.3084, simple_loss=0.3626, pruned_loss=0.1272, over 8466.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3222, pruned_loss=0.08967, over 1585371.43 frames. ], batch size: 25, lr: 8.73e-03, grad_scale: 16.0 +2023-02-06 06:41:17,141 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1201, 1.2534, 1.1407, 0.6278, 1.2484, 0.9734, 0.1536, 1.2131], + device='cuda:1'), covar=tensor([0.0168, 0.0147, 0.0132, 0.0218, 0.0169, 0.0415, 0.0341, 0.0139], + device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0290, 0.0241, 0.0350, 0.0282, 0.0447, 0.0336, 0.0319], + device='cuda:1'), out_proj_covar=tensor([1.0988e-04, 8.4788e-05, 7.1140e-05, 1.0286e-04, 8.4558e-05, 1.4460e-04, + 1.0110e-04, 9.5098e-05], device='cuda:1') +2023-02-06 06:41:25,161 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:41:29,556 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.736e+02 3.271e+02 4.209e+02 1.110e+03, threshold=6.542e+02, percent-clipped=5.0 +2023-02-06 06:41:37,086 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65550.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:41:47,588 INFO [train.py:901] (1/4) Epoch 9, batch 900, loss[loss=0.2416, simple_loss=0.3238, pruned_loss=0.07971, over 8197.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3232, pruned_loss=0.08968, over 1593737.69 frames. ], batch size: 23, lr: 8.73e-03, grad_scale: 16.0 +2023-02-06 06:42:12,150 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7406, 2.4377, 1.8878, 2.0493, 2.0981, 1.6471, 1.9430, 2.0117], + device='cuda:1'), covar=tensor([0.0898, 0.0291, 0.0661, 0.0427, 0.0452, 0.0951, 0.0597, 0.0698], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0237, 0.0312, 0.0300, 0.0304, 0.0320, 0.0341, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 06:42:23,173 INFO [train.py:901] (1/4) Epoch 9, batch 950, loss[loss=0.2351, simple_loss=0.3044, pruned_loss=0.08291, over 8372.00 frames. ], tot_loss[loss=0.251, simple_loss=0.323, pruned_loss=0.08945, over 1598351.92 frames. ], batch size: 24, lr: 8.73e-03, grad_scale: 16.0 +2023-02-06 06:42:39,187 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.498e+02 3.047e+02 4.041e+02 6.463e+02, threshold=6.094e+02, percent-clipped=0.0 +2023-02-06 06:42:44,700 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65647.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:42:50,355 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 06:42:56,309 INFO [train.py:901] (1/4) Epoch 9, batch 1000, loss[loss=0.2553, simple_loss=0.3365, pruned_loss=0.08709, over 8277.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3225, pruned_loss=0.08947, over 1601086.81 frames. ], batch size: 23, lr: 8.72e-03, grad_scale: 16.0 +2023-02-06 06:43:23,104 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 06:43:27,427 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:43:31,856 INFO [train.py:901] (1/4) Epoch 9, batch 1050, loss[loss=0.2271, simple_loss=0.2887, pruned_loss=0.08269, over 7685.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3213, pruned_loss=0.08862, over 1600768.95 frames. ], batch size: 18, lr: 8.72e-03, grad_scale: 16.0 +2023-02-06 06:43:35,709 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 06:43:44,967 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65734.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:43:47,993 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.070e+02 2.877e+02 3.398e+02 4.338e+02 8.070e+02, threshold=6.796e+02, percent-clipped=6.0 +2023-02-06 06:44:00,654 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8031, 3.6989, 3.3995, 1.5347, 3.3679, 3.3798, 3.3847, 3.0744], + device='cuda:1'), covar=tensor([0.1002, 0.0826, 0.1143, 0.5133, 0.0964, 0.1208, 0.1507, 0.1129], + device='cuda:1'), in_proj_covar=tensor([0.0444, 0.0354, 0.0368, 0.0465, 0.0363, 0.0345, 0.0361, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 06:44:03,475 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65762.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:44:05,257 INFO [train.py:901] (1/4) Epoch 9, batch 1100, loss[loss=0.2617, simple_loss=0.3219, pruned_loss=0.1007, over 7973.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3219, pruned_loss=0.08902, over 1605497.21 frames. ], batch size: 21, lr: 8.72e-03, grad_scale: 16.0 +2023-02-06 06:44:20,619 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65788.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:44:38,698 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:44:39,897 INFO [train.py:901] (1/4) Epoch 9, batch 1150, loss[loss=0.258, simple_loss=0.32, pruned_loss=0.09803, over 8236.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.322, pruned_loss=0.08973, over 1607587.03 frames. ], batch size: 22, lr: 8.71e-03, grad_scale: 16.0 +2023-02-06 06:44:44,633 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 06:44:56,767 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.732e+02 2.552e+02 3.121e+02 3.966e+02 8.304e+02, threshold=6.242e+02, percent-clipped=2.0 +2023-02-06 06:45:09,004 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65856.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:45:14,590 INFO [train.py:901] (1/4) Epoch 9, batch 1200, loss[loss=0.2681, simple_loss=0.3354, pruned_loss=0.1003, over 8446.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.323, pruned_loss=0.08995, over 1613402.30 frames. ], batch size: 27, lr: 8.71e-03, grad_scale: 16.0 +2023-02-06 06:45:33,719 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65894.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:45:47,642 INFO [train.py:901] (1/4) Epoch 9, batch 1250, loss[loss=0.2911, simple_loss=0.3529, pruned_loss=0.1147, over 8468.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3228, pruned_loss=0.08992, over 1614334.31 frames. ], batch size: 48, lr: 8.71e-03, grad_scale: 16.0 +2023-02-06 06:46:05,032 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 2.844e+02 3.477e+02 4.312e+02 8.167e+02, threshold=6.953e+02, percent-clipped=5.0 +2023-02-06 06:46:23,818 INFO [train.py:901] (1/4) Epoch 9, batch 1300, loss[loss=0.253, simple_loss=0.3229, pruned_loss=0.09161, over 8638.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3229, pruned_loss=0.09014, over 1616559.21 frames. ], batch size: 34, lr: 8.70e-03, grad_scale: 16.0 +2023-02-06 06:46:25,399 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7025, 1.3386, 1.4989, 1.1819, 0.9308, 1.2871, 1.3750, 1.5745], + device='cuda:1'), covar=tensor([0.0541, 0.1222, 0.1608, 0.1375, 0.0567, 0.1476, 0.0726, 0.0600], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0159, 0.0198, 0.0164, 0.0111, 0.0168, 0.0122, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 06:46:26,118 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5709, 1.9527, 2.0602, 1.2777, 2.2131, 1.4591, 0.7204, 1.7406], + device='cuda:1'), covar=tensor([0.0307, 0.0171, 0.0146, 0.0294, 0.0207, 0.0512, 0.0428, 0.0161], + device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0293, 0.0241, 0.0354, 0.0281, 0.0449, 0.0337, 0.0323], + device='cuda:1'), out_proj_covar=tensor([1.0895e-04, 8.5781e-05, 7.0990e-05, 1.0369e-04, 8.4104e-05, 1.4484e-04, + 1.0126e-04, 9.6215e-05], device='cuda:1') +2023-02-06 06:46:43,342 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7104, 2.6284, 2.8453, 2.0349, 1.5468, 2.8425, 0.7715, 1.9064], + device='cuda:1'), covar=tensor([0.2310, 0.1508, 0.0649, 0.2629, 0.5018, 0.0559, 0.4594, 0.2633], + device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0151, 0.0090, 0.0199, 0.0239, 0.0093, 0.0155, 0.0153], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 06:46:55,288 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66009.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:46:59,203 INFO [train.py:901] (1/4) Epoch 9, batch 1350, loss[loss=0.2737, simple_loss=0.3383, pruned_loss=0.1046, over 8591.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3227, pruned_loss=0.09025, over 1613241.64 frames. ], batch size: 31, lr: 8.70e-03, grad_scale: 8.0 +2023-02-06 06:47:01,384 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:47:17,553 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 2.557e+02 3.336e+02 4.233e+02 1.201e+03, threshold=6.672e+02, percent-clipped=8.0 +2023-02-06 06:47:19,768 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:47:35,102 INFO [train.py:901] (1/4) Epoch 9, batch 1400, loss[loss=0.2299, simple_loss=0.311, pruned_loss=0.0744, over 8132.00 frames. ], tot_loss[loss=0.2517, simple_loss=0.3229, pruned_loss=0.09029, over 1614671.30 frames. ], batch size: 22, lr: 8.70e-03, grad_scale: 8.0 +2023-02-06 06:47:57,581 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8068, 1.7528, 2.4058, 1.3628, 2.0763, 2.6553, 2.7004, 2.1665], + device='cuda:1'), covar=tensor([0.0974, 0.1216, 0.0956, 0.1882, 0.1511, 0.0456, 0.0674, 0.0908], + device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0284, 0.0244, 0.0277, 0.0259, 0.0226, 0.0302, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 06:48:09,442 INFO [train.py:901] (1/4) Epoch 9, batch 1450, loss[loss=0.262, simple_loss=0.3268, pruned_loss=0.09866, over 8612.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3241, pruned_loss=0.09105, over 1615971.78 frames. ], batch size: 34, lr: 8.69e-03, grad_scale: 8.0 +2023-02-06 06:48:12,160 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 06:48:21,933 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 06:48:26,154 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.917e+02 2.633e+02 3.463e+02 4.686e+02 9.003e+02, threshold=6.925e+02, percent-clipped=5.0 +2023-02-06 06:48:42,245 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66162.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:48:44,177 INFO [train.py:901] (1/4) Epoch 9, batch 1500, loss[loss=0.2645, simple_loss=0.336, pruned_loss=0.09653, over 8615.00 frames. ], tot_loss[loss=0.2517, simple_loss=0.3234, pruned_loss=0.09003, over 1619364.61 frames. ], batch size: 31, lr: 8.69e-03, grad_scale: 8.0 +2023-02-06 06:48:52,880 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66178.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:49:08,955 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66200.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:49:18,750 INFO [train.py:901] (1/4) Epoch 9, batch 1550, loss[loss=0.2184, simple_loss=0.2972, pruned_loss=0.06976, over 7968.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3212, pruned_loss=0.08889, over 1616200.58 frames. ], batch size: 21, lr: 8.69e-03, grad_scale: 8.0 +2023-02-06 06:49:35,629 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.559e+02 2.942e+02 3.565e+02 7.942e+02, threshold=5.885e+02, percent-clipped=2.0 +2023-02-06 06:49:53,203 INFO [train.py:901] (1/4) Epoch 9, batch 1600, loss[loss=0.2066, simple_loss=0.2794, pruned_loss=0.06688, over 7710.00 frames. ], tot_loss[loss=0.2514, simple_loss=0.3228, pruned_loss=0.09, over 1620022.98 frames. ], batch size: 18, lr: 8.68e-03, grad_scale: 8.0 +2023-02-06 06:49:53,437 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66265.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:50:02,306 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66276.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:50:11,777 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66290.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:50:29,749 INFO [train.py:901] (1/4) Epoch 9, batch 1650, loss[loss=0.2429, simple_loss=0.3241, pruned_loss=0.08084, over 8201.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3215, pruned_loss=0.08853, over 1618990.20 frames. ], batch size: 23, lr: 8.68e-03, grad_scale: 8.0 +2023-02-06 06:50:29,934 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66315.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:50:46,542 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.749e+02 2.536e+02 3.360e+02 4.258e+02 7.701e+02, threshold=6.719e+02, percent-clipped=5.0 +2023-02-06 06:50:53,612 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 06:51:03,473 INFO [train.py:901] (1/4) Epoch 9, batch 1700, loss[loss=0.2864, simple_loss=0.3577, pruned_loss=0.1075, over 8389.00 frames. ], tot_loss[loss=0.2514, simple_loss=0.3233, pruned_loss=0.08975, over 1621337.82 frames. ], batch size: 49, lr: 8.68e-03, grad_scale: 8.0 +2023-02-06 06:51:12,076 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 06:51:18,054 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4776, 2.8535, 1.8194, 2.1340, 2.1548, 1.4225, 2.1128, 2.0081], + device='cuda:1'), covar=tensor([0.1360, 0.0279, 0.0970, 0.0693, 0.0737, 0.1319, 0.0842, 0.0903], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0232, 0.0312, 0.0299, 0.0305, 0.0318, 0.0341, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 06:51:26,496 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 +2023-02-06 06:51:39,916 INFO [train.py:901] (1/4) Epoch 9, batch 1750, loss[loss=0.2689, simple_loss=0.3551, pruned_loss=0.09141, over 8483.00 frames. ], tot_loss[loss=0.2511, simple_loss=0.3235, pruned_loss=0.08933, over 1623558.63 frames. ], batch size: 27, lr: 8.67e-03, grad_scale: 8.0 +2023-02-06 06:51:53,592 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2802, 1.2581, 1.4827, 1.2396, 0.8116, 1.3005, 1.2371, 1.1312], + device='cuda:1'), covar=tensor([0.0554, 0.1260, 0.1779, 0.1394, 0.0587, 0.1513, 0.0669, 0.0609], + device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0159, 0.0200, 0.0164, 0.0111, 0.0169, 0.0123, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 06:51:57,456 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.817e+02 2.942e+02 3.542e+02 4.261e+02 7.419e+02, threshold=7.084e+02, percent-clipped=2.0 +2023-02-06 06:52:13,916 INFO [train.py:901] (1/4) Epoch 9, batch 1800, loss[loss=0.1988, simple_loss=0.28, pruned_loss=0.05885, over 8094.00 frames. ], tot_loss[loss=0.2511, simple_loss=0.3233, pruned_loss=0.08942, over 1621407.38 frames. ], batch size: 21, lr: 8.67e-03, grad_scale: 8.0 +2023-02-06 06:52:42,187 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66506.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:52:48,995 INFO [train.py:901] (1/4) Epoch 9, batch 1850, loss[loss=0.2508, simple_loss=0.3023, pruned_loss=0.09961, over 7724.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3214, pruned_loss=0.08837, over 1619366.92 frames. ], batch size: 18, lr: 8.67e-03, grad_scale: 8.0 +2023-02-06 06:52:53,698 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66522.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:53:05,495 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66539.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:53:05,989 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.848e+02 3.228e+02 4.154e+02 1.120e+03, threshold=6.457e+02, percent-clipped=1.0 +2023-02-06 06:53:22,984 INFO [train.py:901] (1/4) Epoch 9, batch 1900, loss[loss=0.2106, simple_loss=0.2818, pruned_loss=0.06966, over 7932.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3224, pruned_loss=0.0888, over 1624558.92 frames. ], batch size: 20, lr: 8.66e-03, grad_scale: 8.0 +2023-02-06 06:53:27,127 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66571.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:53:43,854 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66596.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:53:47,139 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 06:53:56,561 INFO [train.py:901] (1/4) Epoch 9, batch 1950, loss[loss=0.2651, simple_loss=0.3406, pruned_loss=0.09478, over 8725.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3229, pruned_loss=0.0891, over 1623294.31 frames. ], batch size: 34, lr: 8.66e-03, grad_scale: 8.0 +2023-02-06 06:53:58,601 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 06:54:00,032 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:54:00,843 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66621.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:54:12,904 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66637.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:54:14,628 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.852e+02 3.410e+02 4.369e+02 9.021e+02, threshold=6.820e+02, percent-clipped=7.0 +2023-02-06 06:54:20,056 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 06:54:32,099 INFO [train.py:901] (1/4) Epoch 9, batch 2000, loss[loss=0.2104, simple_loss=0.2861, pruned_loss=0.06737, over 7811.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.321, pruned_loss=0.08811, over 1619237.25 frames. ], batch size: 20, lr: 8.66e-03, grad_scale: 8.0 +2023-02-06 06:55:06,976 INFO [train.py:901] (1/4) Epoch 9, batch 2050, loss[loss=0.2058, simple_loss=0.2937, pruned_loss=0.059, over 8478.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.321, pruned_loss=0.08808, over 1619963.37 frames. ], batch size: 25, lr: 8.65e-03, grad_scale: 8.0 +2023-02-06 06:55:08,490 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.51 vs. limit=5.0 +2023-02-06 06:55:11,895 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-06 06:55:20,375 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66735.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:55:23,644 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.763e+02 3.349e+02 4.333e+02 1.017e+03, threshold=6.698e+02, percent-clipped=4.0 +2023-02-06 06:55:31,185 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66749.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:55:34,842 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-06 06:55:42,441 INFO [train.py:901] (1/4) Epoch 9, batch 2100, loss[loss=0.2716, simple_loss=0.3406, pruned_loss=0.1013, over 8042.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.3226, pruned_loss=0.08898, over 1620297.70 frames. ], batch size: 22, lr: 8.65e-03, grad_scale: 8.0 +2023-02-06 06:56:17,439 INFO [train.py:901] (1/4) Epoch 9, batch 2150, loss[loss=0.2563, simple_loss=0.3088, pruned_loss=0.1019, over 7532.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3216, pruned_loss=0.08883, over 1616012.20 frames. ], batch size: 18, lr: 8.65e-03, grad_scale: 8.0 +2023-02-06 06:56:25,854 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7284, 1.4644, 1.5377, 1.3038, 1.2700, 1.5297, 2.2044, 2.2984], + device='cuda:1'), covar=tensor([0.0611, 0.1763, 0.2681, 0.1963, 0.0709, 0.2116, 0.0765, 0.0649], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0158, 0.0199, 0.0162, 0.0109, 0.0167, 0.0121, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 06:56:26,603 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 06:56:34,548 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.810e+02 3.362e+02 4.511e+02 1.000e+03, threshold=6.724e+02, percent-clipped=7.0 +2023-02-06 06:56:40,142 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9866, 2.1691, 1.8830, 2.7589, 1.3081, 1.4320, 1.9033, 2.2260], + device='cuda:1'), covar=tensor([0.0831, 0.0943, 0.1075, 0.0449, 0.1342, 0.1653, 0.1028, 0.0822], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0223, 0.0264, 0.0218, 0.0224, 0.0260, 0.0264, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 06:56:49,315 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1879, 2.1836, 1.5713, 2.0144, 1.7467, 1.2728, 1.7520, 1.8660], + device='cuda:1'), covar=tensor([0.1134, 0.0329, 0.1126, 0.0453, 0.0693, 0.1408, 0.0786, 0.0696], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0236, 0.0316, 0.0300, 0.0308, 0.0323, 0.0344, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 06:56:53,078 INFO [train.py:901] (1/4) Epoch 9, batch 2200, loss[loss=0.264, simple_loss=0.3334, pruned_loss=0.09735, over 8196.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3221, pruned_loss=0.08949, over 1615837.16 frames. ], batch size: 23, lr: 8.64e-03, grad_scale: 8.0 +2023-02-06 06:57:01,068 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66877.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:57:05,627 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66883.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:57:05,935 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 06:57:12,258 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66893.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:57:18,133 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66902.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:57:18,322 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 06:57:24,871 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.01 vs. limit=5.0 +2023-02-06 06:57:27,004 INFO [train.py:901] (1/4) Epoch 9, batch 2250, loss[loss=0.2186, simple_loss=0.2839, pruned_loss=0.07665, over 7545.00 frames. ], tot_loss[loss=0.2518, simple_loss=0.3227, pruned_loss=0.0904, over 1614762.70 frames. ], batch size: 18, lr: 8.64e-03, grad_scale: 8.0 +2023-02-06 06:57:29,191 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2584, 1.3767, 1.4360, 1.3053, 0.7652, 1.3764, 1.2274, 1.0168], + device='cuda:1'), covar=tensor([0.0556, 0.1155, 0.1637, 0.1300, 0.0548, 0.1421, 0.0600, 0.0632], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0158, 0.0198, 0.0162, 0.0109, 0.0167, 0.0120, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 06:57:29,219 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66918.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:57:43,696 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.716e+02 3.375e+02 4.203e+02 7.579e+02, threshold=6.750e+02, percent-clipped=1.0 +2023-02-06 06:58:00,326 INFO [train.py:901] (1/4) Epoch 9, batch 2300, loss[loss=0.251, simple_loss=0.3347, pruned_loss=0.08363, over 8337.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.3213, pruned_loss=0.08928, over 1615723.68 frames. ], batch size: 26, lr: 8.64e-03, grad_scale: 8.0 +2023-02-06 06:58:01,250 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0999, 2.2096, 1.6151, 1.9838, 1.7049, 1.3015, 1.6955, 1.7491], + device='cuda:1'), covar=tensor([0.0968, 0.0347, 0.0966, 0.0410, 0.0564, 0.1242, 0.0749, 0.0587], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0236, 0.0314, 0.0299, 0.0305, 0.0319, 0.0342, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 06:58:07,517 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66975.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:58:08,933 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5478, 3.9391, 2.2982, 2.4713, 2.7527, 1.9453, 2.7581, 2.5934], + device='cuda:1'), covar=tensor([0.1488, 0.0223, 0.0809, 0.0774, 0.0617, 0.1143, 0.0838, 0.0975], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0237, 0.0314, 0.0300, 0.0306, 0.0320, 0.0343, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 06:58:19,750 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66991.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:58:24,436 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66998.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:58:36,101 INFO [train.py:901] (1/4) Epoch 9, batch 2350, loss[loss=0.2247, simple_loss=0.3019, pruned_loss=0.07375, over 7978.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3195, pruned_loss=0.08808, over 1615220.71 frames. ], batch size: 21, lr: 8.63e-03, grad_scale: 8.0 +2023-02-06 06:58:36,997 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67016.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:58:53,557 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.755e+02 3.236e+02 4.430e+02 1.005e+03, threshold=6.472e+02, percent-clipped=3.0 +2023-02-06 06:59:10,055 INFO [train.py:901] (1/4) Epoch 9, batch 2400, loss[loss=0.2472, simple_loss=0.3197, pruned_loss=0.08734, over 8235.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3202, pruned_loss=0.08834, over 1619707.28 frames. ], batch size: 22, lr: 8.63e-03, grad_scale: 8.0 +2023-02-06 06:59:28,524 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67093.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:59:34,528 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67101.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:59:44,443 INFO [train.py:901] (1/4) Epoch 9, batch 2450, loss[loss=0.2341, simple_loss=0.311, pruned_loss=0.07856, over 8344.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3188, pruned_loss=0.08725, over 1617129.36 frames. ], batch size: 26, lr: 8.63e-03, grad_scale: 8.0 +2023-02-06 07:00:02,010 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.787e+02 3.467e+02 4.148e+02 8.119e+02, threshold=6.934e+02, percent-clipped=3.0 +2023-02-06 07:00:18,498 INFO [train.py:901] (1/4) Epoch 9, batch 2500, loss[loss=0.2318, simple_loss=0.3163, pruned_loss=0.07365, over 8446.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.3214, pruned_loss=0.0892, over 1619948.94 frames. ], batch size: 27, lr: 8.62e-03, grad_scale: 8.0 +2023-02-06 07:00:18,669 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1306, 1.6603, 1.5490, 1.3065, 1.0983, 1.3623, 1.7329, 1.7647], + device='cuda:1'), covar=tensor([0.0526, 0.1136, 0.1749, 0.1391, 0.0619, 0.1539, 0.0719, 0.0548], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0157, 0.0196, 0.0161, 0.0108, 0.0167, 0.0120, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 07:00:23,019 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.32 vs. limit=5.0 +2023-02-06 07:00:48,186 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67208.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:00:52,871 INFO [train.py:901] (1/4) Epoch 9, batch 2550, loss[loss=0.2339, simple_loss=0.3221, pruned_loss=0.07288, over 8480.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3217, pruned_loss=0.08885, over 1620045.67 frames. ], batch size: 29, lr: 8.62e-03, grad_scale: 8.0 +2023-02-06 07:01:12,346 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.679e+02 3.405e+02 4.203e+02 8.726e+02, threshold=6.810e+02, percent-clipped=2.0 +2023-02-06 07:01:21,884 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67254.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:01:29,617 INFO [train.py:901] (1/4) Epoch 9, batch 2600, loss[loss=0.2453, simple_loss=0.3092, pruned_loss=0.09069, over 7520.00 frames. ], tot_loss[loss=0.2488, simple_loss=0.3208, pruned_loss=0.0884, over 1617224.08 frames. ], batch size: 18, lr: 8.62e-03, grad_scale: 8.0 +2023-02-06 07:01:36,607 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5405, 1.5884, 4.4155, 1.9962, 2.4216, 5.0379, 4.9651, 4.2906], + device='cuda:1'), covar=tensor([0.1073, 0.1714, 0.0255, 0.1946, 0.1007, 0.0184, 0.0346, 0.0591], + device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0281, 0.0244, 0.0274, 0.0256, 0.0227, 0.0302, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:01:39,252 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67279.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:02:03,430 INFO [train.py:901] (1/4) Epoch 9, batch 2650, loss[loss=0.2276, simple_loss=0.2975, pruned_loss=0.07889, over 7806.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.32, pruned_loss=0.08812, over 1616745.14 frames. ], batch size: 19, lr: 8.62e-03, grad_scale: 8.0 +2023-02-06 07:02:06,282 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67319.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:02:11,466 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.02 vs. limit=5.0 +2023-02-06 07:02:19,993 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2828, 1.8219, 2.8709, 2.2572, 2.4082, 2.1166, 1.5590, 1.1379], + device='cuda:1'), covar=tensor([0.3294, 0.3555, 0.0909, 0.2137, 0.1727, 0.1888, 0.1620, 0.3663], + device='cuda:1'), in_proj_covar=tensor([0.0853, 0.0818, 0.0698, 0.0810, 0.0900, 0.0761, 0.0686, 0.0732], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 07:02:21,820 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.856e+02 2.739e+02 3.376e+02 4.238e+02 9.756e+02, threshold=6.752e+02, percent-clipped=4.0 +2023-02-06 07:02:38,172 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4931, 1.4455, 1.6735, 1.3371, 0.9781, 1.7097, 0.1651, 1.1645], + device='cuda:1'), covar=tensor([0.2852, 0.1957, 0.0737, 0.2119, 0.5446, 0.0638, 0.4027, 0.2194], + device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0154, 0.0093, 0.0205, 0.0244, 0.0095, 0.0154, 0.0155], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 07:02:39,314 INFO [train.py:901] (1/4) Epoch 9, batch 2700, loss[loss=0.2266, simple_loss=0.3022, pruned_loss=0.07545, over 8133.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3195, pruned_loss=0.08771, over 1617592.75 frames. ], batch size: 22, lr: 8.61e-03, grad_scale: 8.0 +2023-02-06 07:02:49,626 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3744, 1.6845, 1.7206, 1.0095, 1.7389, 1.2928, 0.3099, 1.5973], + device='cuda:1'), covar=tensor([0.0342, 0.0186, 0.0214, 0.0292, 0.0311, 0.0620, 0.0534, 0.0160], + device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0294, 0.0246, 0.0359, 0.0285, 0.0446, 0.0339, 0.0323], + device='cuda:1'), out_proj_covar=tensor([1.1038e-04, 8.5350e-05, 7.2293e-05, 1.0518e-04, 8.4636e-05, 1.4294e-04, + 1.0141e-04, 9.5693e-05], device='cuda:1') +2023-02-06 07:03:13,103 INFO [train.py:901] (1/4) Epoch 9, batch 2750, loss[loss=0.2753, simple_loss=0.3345, pruned_loss=0.1081, over 7525.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3195, pruned_loss=0.08831, over 1614183.70 frames. ], batch size: 18, lr: 8.61e-03, grad_scale: 8.0 +2023-02-06 07:03:15,954 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1557, 1.2845, 4.2290, 1.6847, 2.3039, 4.7856, 4.8936, 3.8464], + device='cuda:1'), covar=tensor([0.1224, 0.1880, 0.0347, 0.2108, 0.1056, 0.0290, 0.0449, 0.0851], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0283, 0.0244, 0.0276, 0.0257, 0.0226, 0.0303, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:03:26,091 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:03:29,964 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.803e+02 2.867e+02 3.446e+02 4.196e+02 9.783e+02, threshold=6.892e+02, percent-clipped=3.0 +2023-02-06 07:03:34,270 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67445.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:03:48,020 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67464.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:03:48,380 INFO [train.py:901] (1/4) Epoch 9, batch 2800, loss[loss=0.3119, simple_loss=0.3748, pruned_loss=0.1245, over 8454.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3208, pruned_loss=0.08911, over 1614619.18 frames. ], batch size: 27, lr: 8.61e-03, grad_scale: 8.0 +2023-02-06 07:04:05,350 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67489.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:04:23,689 INFO [train.py:901] (1/4) Epoch 9, batch 2850, loss[loss=0.2238, simple_loss=0.3158, pruned_loss=0.0659, over 8322.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.321, pruned_loss=0.08943, over 1614768.03 frames. ], batch size: 25, lr: 8.60e-03, grad_scale: 8.0 +2023-02-06 07:04:34,554 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:04:40,481 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.340e+02 2.765e+02 3.269e+02 4.105e+02 6.649e+02, threshold=6.538e+02, percent-clipped=0.0 +2023-02-06 07:04:55,076 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67560.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:04:58,183 INFO [train.py:901] (1/4) Epoch 9, batch 2900, loss[loss=0.2192, simple_loss=0.2892, pruned_loss=0.07466, over 7505.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3209, pruned_loss=0.08896, over 1618267.05 frames. ], batch size: 18, lr: 8.60e-03, grad_scale: 8.0 +2023-02-06 07:05:02,203 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-06 07:05:24,258 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 07:05:33,887 INFO [train.py:901] (1/4) Epoch 9, batch 2950, loss[loss=0.2514, simple_loss=0.323, pruned_loss=0.08988, over 8367.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3208, pruned_loss=0.08831, over 1616813.02 frames. ], batch size: 48, lr: 8.60e-03, grad_scale: 8.0 +2023-02-06 07:05:39,065 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2621, 2.2089, 1.4953, 1.9384, 1.8631, 1.2809, 1.5774, 1.8100], + device='cuda:1'), covar=tensor([0.1029, 0.0291, 0.1008, 0.0440, 0.0530, 0.1240, 0.0861, 0.0760], + device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0232, 0.0311, 0.0296, 0.0303, 0.0317, 0.0340, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 07:05:51,259 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.115e+02 2.827e+02 3.390e+02 4.435e+02 7.404e+02, threshold=6.780e+02, percent-clipped=4.0 +2023-02-06 07:06:08,216 INFO [train.py:901] (1/4) Epoch 9, batch 3000, loss[loss=0.263, simple_loss=0.3413, pruned_loss=0.09231, over 8322.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.321, pruned_loss=0.08772, over 1621039.05 frames. ], batch size: 25, lr: 8.59e-03, grad_scale: 8.0 +2023-02-06 07:06:08,217 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 07:06:20,341 INFO [train.py:935] (1/4) Epoch 9, validation: loss=0.1965, simple_loss=0.2957, pruned_loss=0.04864, over 944034.00 frames. +2023-02-06 07:06:20,342 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 07:06:37,422 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67690.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:06:43,349 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:06:52,180 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67710.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:06:55,505 INFO [train.py:901] (1/4) Epoch 9, batch 3050, loss[loss=0.2281, simple_loss=0.2962, pruned_loss=0.07998, over 7249.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3209, pruned_loss=0.08784, over 1617491.77 frames. ], batch size: 16, lr: 8.59e-03, grad_scale: 8.0 +2023-02-06 07:06:55,723 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:07:13,238 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.763e+02 2.629e+02 3.194e+02 3.976e+02 7.575e+02, threshold=6.387e+02, percent-clipped=1.0 +2023-02-06 07:07:29,777 INFO [train.py:901] (1/4) Epoch 9, batch 3100, loss[loss=0.2467, simple_loss=0.325, pruned_loss=0.08422, over 8503.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3199, pruned_loss=0.08726, over 1617340.93 frames. ], batch size: 28, lr: 8.59e-03, grad_scale: 8.0 +2023-02-06 07:07:52,369 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([6.0155, 6.1261, 5.2382, 2.3677, 5.2254, 5.6342, 5.5467, 5.3106], + device='cuda:1'), covar=tensor([0.0565, 0.0366, 0.0802, 0.4252, 0.0682, 0.0638, 0.1083, 0.0634], + device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0343, 0.0361, 0.0447, 0.0356, 0.0335, 0.0354, 0.0299], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 07:08:04,172 INFO [train.py:901] (1/4) Epoch 9, batch 3150, loss[loss=0.2209, simple_loss=0.3195, pruned_loss=0.06117, over 8509.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3217, pruned_loss=0.08822, over 1619837.18 frames. ], batch size: 26, lr: 8.58e-03, grad_scale: 8.0 +2023-02-06 07:08:05,040 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67816.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:08:21,147 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.202e+02 2.768e+02 3.401e+02 4.235e+02 8.418e+02, threshold=6.801e+02, percent-clipped=5.0 +2023-02-06 07:08:21,987 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:08:37,988 INFO [train.py:901] (1/4) Epoch 9, batch 3200, loss[loss=0.2979, simple_loss=0.3507, pruned_loss=0.1226, over 8524.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3228, pruned_loss=0.08902, over 1621085.57 frames. ], batch size: 28, lr: 8.58e-03, grad_scale: 8.0 +2023-02-06 07:08:42,800 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2037, 1.4267, 1.5113, 1.3863, 1.0943, 1.3697, 1.6192, 1.5598], + device='cuda:1'), covar=tensor([0.0470, 0.1238, 0.1703, 0.1318, 0.0564, 0.1482, 0.0721, 0.0590], + device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0158, 0.0197, 0.0162, 0.0110, 0.0168, 0.0121, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 07:08:44,730 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:09:12,191 INFO [train.py:901] (1/4) Epoch 9, batch 3250, loss[loss=0.3532, simple_loss=0.414, pruned_loss=0.1462, over 8481.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3216, pruned_loss=0.08881, over 1618192.29 frames. ], batch size: 27, lr: 8.58e-03, grad_scale: 8.0 +2023-02-06 07:09:29,470 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.709e+02 3.362e+02 4.203e+02 8.128e+02, threshold=6.724e+02, percent-clipped=5.0 +2023-02-06 07:09:46,666 INFO [train.py:901] (1/4) Epoch 9, batch 3300, loss[loss=0.2378, simple_loss=0.3074, pruned_loss=0.08411, over 7964.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3218, pruned_loss=0.08909, over 1617912.72 frames. ], batch size: 21, lr: 8.57e-03, grad_scale: 8.0 +2023-02-06 07:09:50,091 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7070, 5.8995, 5.0095, 2.0684, 5.0270, 5.5481, 5.4591, 4.9927], + device='cuda:1'), covar=tensor([0.0525, 0.0338, 0.0737, 0.4553, 0.0648, 0.0576, 0.0700, 0.0689], + device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0347, 0.0360, 0.0454, 0.0360, 0.0340, 0.0356, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 07:10:04,401 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67990.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:10:10,457 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:10:22,541 INFO [train.py:901] (1/4) Epoch 9, batch 3350, loss[loss=0.2443, simple_loss=0.3201, pruned_loss=0.08427, over 8322.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3219, pruned_loss=0.08903, over 1616526.27 frames. ], batch size: 25, lr: 8.57e-03, grad_scale: 16.0 +2023-02-06 07:10:39,231 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 2.493e+02 3.108e+02 4.287e+02 1.101e+03, threshold=6.217e+02, percent-clipped=5.0 +2023-02-06 07:10:40,592 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:10:46,026 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.66 vs. limit=5.0 +2023-02-06 07:10:49,087 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68054.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:10:56,401 INFO [train.py:901] (1/4) Epoch 9, batch 3400, loss[loss=0.2524, simple_loss=0.3246, pruned_loss=0.09007, over 8359.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3215, pruned_loss=0.08857, over 1618852.90 frames. ], batch size: 24, lr: 8.57e-03, grad_scale: 16.0 +2023-02-06 07:11:01,258 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 07:11:24,317 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68105.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:11:25,610 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5224, 2.9929, 1.8140, 2.2137, 2.3860, 1.4828, 1.9977, 2.1609], + device='cuda:1'), covar=tensor([0.1367, 0.0288, 0.0985, 0.0653, 0.0587, 0.1400, 0.0976, 0.0987], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0234, 0.0310, 0.0295, 0.0305, 0.0320, 0.0342, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 07:11:30,827 INFO [train.py:901] (1/4) Epoch 9, batch 3450, loss[loss=0.2885, simple_loss=0.36, pruned_loss=0.1085, over 8195.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3219, pruned_loss=0.08865, over 1622834.03 frames. ], batch size: 23, lr: 8.56e-03, grad_scale: 16.0 +2023-02-06 07:11:48,136 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.585e+02 3.242e+02 3.955e+02 1.617e+03, threshold=6.484e+02, percent-clipped=7.0 +2023-02-06 07:11:59,844 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68157.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:12:05,792 INFO [train.py:901] (1/4) Epoch 9, batch 3500, loss[loss=0.2945, simple_loss=0.3673, pruned_loss=0.1109, over 8513.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3229, pruned_loss=0.0891, over 1623119.41 frames. ], batch size: 28, lr: 8.56e-03, grad_scale: 8.0 +2023-02-06 07:12:08,693 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:12:17,615 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.09 vs. limit=5.0 +2023-02-06 07:12:17,936 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 07:12:40,958 INFO [train.py:901] (1/4) Epoch 9, batch 3550, loss[loss=0.2104, simple_loss=0.2774, pruned_loss=0.07169, over 7930.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3218, pruned_loss=0.08848, over 1620546.54 frames. ], batch size: 20, lr: 8.56e-03, grad_scale: 8.0 +2023-02-06 07:12:48,010 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.23 vs. limit=2.0 +2023-02-06 07:12:58,946 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.838e+02 3.387e+02 4.304e+02 7.616e+02, threshold=6.774e+02, percent-clipped=6.0 +2023-02-06 07:13:02,583 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68246.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:13:14,900 INFO [train.py:901] (1/4) Epoch 9, batch 3600, loss[loss=0.3091, simple_loss=0.3682, pruned_loss=0.125, over 8492.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3231, pruned_loss=0.08949, over 1620115.11 frames. ], batch size: 28, lr: 8.56e-03, grad_scale: 8.0 +2023-02-06 07:13:19,091 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68271.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:13:49,592 INFO [train.py:901] (1/4) Epoch 9, batch 3650, loss[loss=0.2302, simple_loss=0.2967, pruned_loss=0.08186, over 7650.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.3226, pruned_loss=0.08926, over 1621565.11 frames. ], batch size: 19, lr: 8.55e-03, grad_scale: 8.0 +2023-02-06 07:14:08,229 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.637e+02 3.214e+02 4.100e+02 7.421e+02, threshold=6.428e+02, percent-clipped=2.0 +2023-02-06 07:14:09,681 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:14:18,271 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 07:14:23,153 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3973, 2.5686, 1.7773, 2.1814, 2.0937, 1.5160, 1.8141, 1.9824], + device='cuda:1'), covar=tensor([0.1165, 0.0335, 0.0853, 0.0471, 0.0595, 0.1197, 0.0886, 0.0708], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0234, 0.0312, 0.0297, 0.0308, 0.0324, 0.0345, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 07:14:25,007 INFO [train.py:901] (1/4) Epoch 9, batch 3700, loss[loss=0.2108, simple_loss=0.2942, pruned_loss=0.06369, over 7973.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3206, pruned_loss=0.08834, over 1615455.67 frames. ], batch size: 21, lr: 8.55e-03, grad_scale: 8.0 +2023-02-06 07:14:40,538 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0031, 2.2944, 3.7908, 2.9393, 3.2260, 2.5484, 2.0345, 1.9349], + device='cuda:1'), covar=tensor([0.3001, 0.3725, 0.0882, 0.2090, 0.1778, 0.1785, 0.1452, 0.3712], + device='cuda:1'), in_proj_covar=tensor([0.0852, 0.0808, 0.0686, 0.0802, 0.0898, 0.0754, 0.0682, 0.0731], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 07:14:57,592 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68413.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:14:58,734 INFO [train.py:901] (1/4) Epoch 9, batch 3750, loss[loss=0.2467, simple_loss=0.3054, pruned_loss=0.09396, over 7232.00 frames. ], tot_loss[loss=0.2488, simple_loss=0.3204, pruned_loss=0.08857, over 1611503.15 frames. ], batch size: 16, lr: 8.55e-03, grad_scale: 8.0 +2023-02-06 07:15:00,178 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:06,051 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68425.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:14,922 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:16,805 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.868e+02 3.639e+02 4.960e+02 1.282e+03, threshold=7.278e+02, percent-clipped=8.0 +2023-02-06 07:15:22,903 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68449.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:15:23,652 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68450.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:29,219 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68458.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:33,828 INFO [train.py:901] (1/4) Epoch 9, batch 3800, loss[loss=0.2707, simple_loss=0.3337, pruned_loss=0.1038, over 8607.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3215, pruned_loss=0.08926, over 1618773.90 frames. ], batch size: 31, lr: 8.54e-03, grad_scale: 8.0 +2023-02-06 07:16:07,820 INFO [train.py:901] (1/4) Epoch 9, batch 3850, loss[loss=0.2053, simple_loss=0.2701, pruned_loss=0.07023, over 7695.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3218, pruned_loss=0.08985, over 1616367.03 frames. ], batch size: 18, lr: 8.54e-03, grad_scale: 8.0 +2023-02-06 07:16:15,346 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5439, 1.8457, 2.0989, 1.2364, 2.2140, 1.4500, 0.7733, 1.7195], + device='cuda:1'), covar=tensor([0.0408, 0.0197, 0.0138, 0.0346, 0.0203, 0.0518, 0.0492, 0.0197], + device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0295, 0.0249, 0.0358, 0.0288, 0.0445, 0.0340, 0.0321], + device='cuda:1'), out_proj_covar=tensor([1.1072e-04, 8.5281e-05, 7.3058e-05, 1.0468e-04, 8.5559e-05, 1.4219e-04, + 1.0140e-04, 9.4757e-05], device='cuda:1') +2023-02-06 07:16:25,012 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 07:16:25,654 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.582e+02 3.048e+02 3.724e+02 6.674e+02, threshold=6.096e+02, percent-clipped=0.0 +2023-02-06 07:16:27,990 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0621, 1.3396, 4.2635, 1.5119, 3.6718, 3.5588, 3.8087, 3.6847], + device='cuda:1'), covar=tensor([0.0532, 0.4276, 0.0468, 0.3450, 0.1164, 0.0838, 0.0526, 0.0622], + device='cuda:1'), in_proj_covar=tensor([0.0443, 0.0546, 0.0542, 0.0509, 0.0574, 0.0479, 0.0476, 0.0539], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 07:16:42,279 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68564.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:16:42,694 INFO [train.py:901] (1/4) Epoch 9, batch 3900, loss[loss=0.2746, simple_loss=0.3354, pruned_loss=0.1069, over 7644.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3216, pruned_loss=0.08961, over 1619235.47 frames. ], batch size: 19, lr: 8.54e-03, grad_scale: 8.0 +2023-02-06 07:17:02,946 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1648, 2.1411, 2.8471, 1.7328, 2.4030, 3.1742, 3.0662, 2.8430], + device='cuda:1'), covar=tensor([0.0787, 0.0976, 0.0607, 0.1559, 0.1238, 0.0259, 0.0562, 0.0491], + device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0285, 0.0246, 0.0276, 0.0262, 0.0228, 0.0304, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:17:04,304 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68596.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:17:17,702 INFO [train.py:901] (1/4) Epoch 9, batch 3950, loss[loss=0.2617, simple_loss=0.3421, pruned_loss=0.0907, over 8576.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.3216, pruned_loss=0.08977, over 1615740.09 frames. ], batch size: 39, lr: 8.53e-03, grad_scale: 8.0 +2023-02-06 07:17:35,330 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.589e+02 3.045e+02 4.133e+02 1.084e+03, threshold=6.090e+02, percent-clipped=3.0 +2023-02-06 07:17:51,773 INFO [train.py:901] (1/4) Epoch 9, batch 4000, loss[loss=0.2476, simple_loss=0.3226, pruned_loss=0.08625, over 8100.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3215, pruned_loss=0.08999, over 1611278.50 frames. ], batch size: 23, lr: 8.53e-03, grad_scale: 8.0 +2023-02-06 07:18:00,059 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.60 vs. limit=5.0 +2023-02-06 07:18:13,682 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.07 vs. limit=5.0 +2023-02-06 07:18:16,919 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-06 07:18:17,936 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68703.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:18:25,237 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68714.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:18:25,676 INFO [train.py:901] (1/4) Epoch 9, batch 4050, loss[loss=0.258, simple_loss=0.3284, pruned_loss=0.09385, over 7981.00 frames. ], tot_loss[loss=0.2511, simple_loss=0.322, pruned_loss=0.09012, over 1613236.06 frames. ], batch size: 21, lr: 8.53e-03, grad_scale: 8.0 +2023-02-06 07:18:28,563 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0097, 2.2274, 4.0318, 2.7704, 3.2876, 2.4204, 1.9691, 1.6518], + device='cuda:1'), covar=tensor([0.3097, 0.3996, 0.0794, 0.2268, 0.1770, 0.2145, 0.1641, 0.4059], + device='cuda:1'), in_proj_covar=tensor([0.0853, 0.0814, 0.0697, 0.0808, 0.0904, 0.0758, 0.0687, 0.0731], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 07:18:42,617 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68739.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:18:43,696 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.898e+02 2.478e+02 3.133e+02 3.692e+02 8.585e+02, threshold=6.266e+02, percent-clipped=3.0 +2023-02-06 07:18:57,742 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:19:00,369 INFO [train.py:901] (1/4) Epoch 9, batch 4100, loss[loss=0.2091, simple_loss=0.286, pruned_loss=0.06614, over 7647.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3211, pruned_loss=0.0889, over 1611661.47 frames. ], batch size: 19, lr: 8.52e-03, grad_scale: 8.0 +2023-02-06 07:19:34,811 INFO [train.py:901] (1/4) Epoch 9, batch 4150, loss[loss=0.2263, simple_loss=0.2999, pruned_loss=0.07637, over 7526.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3206, pruned_loss=0.08899, over 1609818.04 frames. ], batch size: 18, lr: 8.52e-03, grad_scale: 8.0 +2023-02-06 07:19:38,351 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68820.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:19:52,100 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 2.558e+02 3.576e+02 4.352e+02 8.740e+02, threshold=7.151e+02, percent-clipped=5.0 +2023-02-06 07:19:55,822 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68845.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:20:02,868 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 07:20:09,494 INFO [train.py:901] (1/4) Epoch 9, batch 4200, loss[loss=0.2268, simple_loss=0.2955, pruned_loss=0.07912, over 7690.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3203, pruned_loss=0.08881, over 1612872.95 frames. ], batch size: 18, lr: 8.52e-03, grad_scale: 8.0 +2023-02-06 07:20:12,586 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-06 07:20:17,185 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:20:23,114 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 07:20:44,838 INFO [train.py:901] (1/4) Epoch 9, batch 4250, loss[loss=0.2269, simple_loss=0.3007, pruned_loss=0.07652, over 7222.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3194, pruned_loss=0.08868, over 1609163.02 frames. ], batch size: 16, lr: 8.52e-03, grad_scale: 8.0 +2023-02-06 07:20:45,559 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 07:21:02,845 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68940.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:21:03,455 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.856e+02 3.701e+02 4.402e+02 9.379e+02, threshold=7.403e+02, percent-clipped=2.0 +2023-02-06 07:21:20,740 INFO [train.py:901] (1/4) Epoch 9, batch 4300, loss[loss=0.2204, simple_loss=0.2901, pruned_loss=0.07541, over 7670.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.32, pruned_loss=0.08909, over 1613407.08 frames. ], batch size: 19, lr: 8.51e-03, grad_scale: 8.0 +2023-02-06 07:21:55,039 INFO [train.py:901] (1/4) Epoch 9, batch 4350, loss[loss=0.2498, simple_loss=0.3133, pruned_loss=0.09315, over 7931.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3218, pruned_loss=0.09011, over 1618415.80 frames. ], batch size: 20, lr: 8.51e-03, grad_scale: 8.0 +2023-02-06 07:22:12,980 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.761e+02 3.203e+02 3.985e+02 6.558e+02, threshold=6.405e+02, percent-clipped=0.0 +2023-02-06 07:22:16,277 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 07:22:17,703 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69047.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:22:23,136 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:22:27,142 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69061.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:22:29,683 INFO [train.py:901] (1/4) Epoch 9, batch 4400, loss[loss=0.2909, simple_loss=0.3614, pruned_loss=0.1101, over 8427.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3222, pruned_loss=0.09046, over 1616449.43 frames. ], batch size: 49, lr: 8.51e-03, grad_scale: 8.0 +2023-02-06 07:22:30,553 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2959, 1.4093, 1.2096, 1.8665, 0.7079, 1.0841, 1.2022, 1.4322], + device='cuda:1'), covar=tensor([0.1038, 0.0925, 0.1250, 0.0569, 0.1353, 0.1703, 0.1026, 0.0829], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0226, 0.0267, 0.0222, 0.0228, 0.0263, 0.0268, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 07:22:40,715 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0288, 1.9278, 2.8637, 1.7440, 2.3850, 3.1309, 3.0661, 2.7035], + device='cuda:1'), covar=tensor([0.0839, 0.1143, 0.0700, 0.1536, 0.1283, 0.0299, 0.0599, 0.0562], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0285, 0.0246, 0.0275, 0.0260, 0.0229, 0.0301, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:22:55,799 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 07:23:04,531 INFO [train.py:901] (1/4) Epoch 9, batch 4450, loss[loss=0.2095, simple_loss=0.2786, pruned_loss=0.07016, over 7533.00 frames. ], tot_loss[loss=0.2502, simple_loss=0.3211, pruned_loss=0.08966, over 1612911.39 frames. ], batch size: 18, lr: 8.50e-03, grad_scale: 8.0 +2023-02-06 07:23:16,661 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69132.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:23:22,374 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.746e+02 3.298e+02 3.852e+02 8.052e+02, threshold=6.596e+02, percent-clipped=4.0 +2023-02-06 07:23:33,872 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69157.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:23:37,258 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:23:39,014 INFO [train.py:901] (1/4) Epoch 9, batch 4500, loss[loss=0.2294, simple_loss=0.3192, pruned_loss=0.06977, over 8454.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3213, pruned_loss=0.08937, over 1613393.65 frames. ], batch size: 27, lr: 8.50e-03, grad_scale: 8.0 +2023-02-06 07:23:49,135 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 07:23:49,292 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69180.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:23:53,178 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69186.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:24:13,193 INFO [train.py:901] (1/4) Epoch 9, batch 4550, loss[loss=0.2403, simple_loss=0.3135, pruned_loss=0.08357, over 8356.00 frames. ], tot_loss[loss=0.2488, simple_loss=0.3202, pruned_loss=0.08871, over 1609067.82 frames. ], batch size: 24, lr: 8.50e-03, grad_scale: 8.0 +2023-02-06 07:24:31,950 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 2.522e+02 2.943e+02 3.743e+02 5.945e+02, threshold=5.886e+02, percent-clipped=0.0 +2023-02-06 07:24:33,347 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69243.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:24:33,984 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0839, 1.7897, 3.7461, 1.6437, 2.5203, 4.0219, 4.2867, 3.1446], + device='cuda:1'), covar=tensor([0.1446, 0.1781, 0.0488, 0.2493, 0.1128, 0.0471, 0.0453, 0.1110], + device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0290, 0.0252, 0.0283, 0.0267, 0.0236, 0.0310, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:24:47,713 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69263.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:24:48,927 INFO [train.py:901] (1/4) Epoch 9, batch 4600, loss[loss=0.2116, simple_loss=0.2796, pruned_loss=0.07184, over 7172.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3201, pruned_loss=0.0885, over 1609850.49 frames. ], batch size: 16, lr: 8.49e-03, grad_scale: 8.0 +2023-02-06 07:25:07,550 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:25:22,106 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69311.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:25:24,593 INFO [train.py:901] (1/4) Epoch 9, batch 4650, loss[loss=0.2735, simple_loss=0.3501, pruned_loss=0.09847, over 8245.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3196, pruned_loss=0.08747, over 1612755.19 frames. ], batch size: 24, lr: 8.49e-03, grad_scale: 8.0 +2023-02-06 07:25:38,787 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:25:40,114 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:25:42,602 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.666e+02 3.298e+02 3.900e+02 8.712e+02, threshold=6.595e+02, percent-clipped=8.0 +2023-02-06 07:25:58,525 INFO [train.py:901] (1/4) Epoch 9, batch 4700, loss[loss=0.228, simple_loss=0.2937, pruned_loss=0.08114, over 7698.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3193, pruned_loss=0.08767, over 1612770.94 frames. ], batch size: 18, lr: 8.49e-03, grad_scale: 8.0 +2023-02-06 07:26:14,067 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69386.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:26:26,888 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69405.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:26:31,849 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69412.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:26:33,682 INFO [train.py:901] (1/4) Epoch 9, batch 4750, loss[loss=0.227, simple_loss=0.3005, pruned_loss=0.07676, over 8292.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3187, pruned_loss=0.0873, over 1614068.23 frames. ], batch size: 23, lr: 8.48e-03, grad_scale: 8.0 +2023-02-06 07:26:35,879 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69418.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:26:43,651 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-06 07:26:44,850 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 07:26:47,353 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0585, 2.6842, 2.9543, 1.2454, 3.0446, 1.9503, 1.3470, 1.9741], + device='cuda:1'), covar=tensor([0.0509, 0.0187, 0.0167, 0.0449, 0.0270, 0.0502, 0.0528, 0.0255], + device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0297, 0.0249, 0.0360, 0.0289, 0.0448, 0.0340, 0.0323], + device='cuda:1'), out_proj_covar=tensor([1.1117e-04, 8.5849e-05, 7.2697e-05, 1.0501e-04, 8.5473e-05, 1.4261e-04, + 1.0118e-04, 9.5352e-05], device='cuda:1') +2023-02-06 07:26:49,139 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 07:26:50,997 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 07:26:51,649 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.564e+02 3.173e+02 4.227e+02 9.736e+02, threshold=6.346e+02, percent-clipped=4.0 +2023-02-06 07:26:53,215 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69443.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:27:08,536 INFO [train.py:901] (1/4) Epoch 9, batch 4800, loss[loss=0.2147, simple_loss=0.3065, pruned_loss=0.06151, over 8500.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3203, pruned_loss=0.08815, over 1618814.69 frames. ], batch size: 26, lr: 8.48e-03, grad_scale: 8.0 +2023-02-06 07:27:38,647 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4166, 1.1903, 4.6710, 1.7787, 3.9899, 3.8103, 4.1583, 4.0612], + device='cuda:1'), covar=tensor([0.0687, 0.4552, 0.0451, 0.3144, 0.1171, 0.0814, 0.0569, 0.0683], + device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0549, 0.0545, 0.0505, 0.0576, 0.0488, 0.0481, 0.0547], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 07:27:41,292 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 07:27:43,207 INFO [train.py:901] (1/4) Epoch 9, batch 4850, loss[loss=0.24, simple_loss=0.311, pruned_loss=0.08449, over 7970.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3211, pruned_loss=0.08849, over 1620792.96 frames. ], batch size: 21, lr: 8.48e-03, grad_scale: 8.0 +2023-02-06 07:27:46,750 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:27:49,410 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:27:53,332 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69530.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:27:55,750 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-06 07:28:00,630 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.717e+02 3.193e+02 3.973e+02 8.915e+02, threshold=6.387e+02, percent-clipped=1.0 +2023-02-06 07:28:17,597 INFO [train.py:901] (1/4) Epoch 9, batch 4900, loss[loss=0.2555, simple_loss=0.3326, pruned_loss=0.08921, over 8249.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3209, pruned_loss=0.0887, over 1617074.06 frames. ], batch size: 24, lr: 8.48e-03, grad_scale: 8.0 +2023-02-06 07:28:33,124 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:28:47,471 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69607.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:28:53,430 INFO [train.py:901] (1/4) Epoch 9, batch 4950, loss[loss=0.2261, simple_loss=0.311, pruned_loss=0.07064, over 8765.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3197, pruned_loss=0.08855, over 1615971.20 frames. ], batch size: 30, lr: 8.47e-03, grad_scale: 8.0 +2023-02-06 07:29:06,609 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:29:09,357 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69639.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:29:10,455 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.734e+02 3.225e+02 4.131e+02 8.295e+02, threshold=6.450e+02, percent-clipped=5.0 +2023-02-06 07:29:13,216 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:29:27,256 INFO [train.py:901] (1/4) Epoch 9, batch 5000, loss[loss=0.2777, simple_loss=0.3279, pruned_loss=0.1138, over 7924.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3193, pruned_loss=0.08837, over 1615639.68 frames. ], batch size: 20, lr: 8.47e-03, grad_scale: 8.0 +2023-02-06 07:29:39,118 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69682.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:29:46,885 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69692.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:29:53,587 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69702.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:02,818 INFO [train.py:901] (1/4) Epoch 9, batch 5050, loss[loss=0.253, simple_loss=0.327, pruned_loss=0.08949, over 8465.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3194, pruned_loss=0.08803, over 1618072.01 frames. ], batch size: 25, lr: 8.47e-03, grad_scale: 8.0 +2023-02-06 07:30:07,759 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69722.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:30:12,767 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69730.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:18,111 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 07:30:20,811 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.703e+02 3.249e+02 3.895e+02 8.845e+02, threshold=6.498e+02, percent-clipped=2.0 +2023-02-06 07:30:26,982 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:30,972 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69756.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:36,716 INFO [train.py:901] (1/4) Epoch 9, batch 5100, loss[loss=0.2125, simple_loss=0.2968, pruned_loss=0.06406, over 8289.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3212, pruned_loss=0.0887, over 1622468.60 frames. ], batch size: 23, lr: 8.46e-03, grad_scale: 8.0 +2023-02-06 07:30:44,225 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69776.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:59,038 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69797.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:31:01,898 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69801.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:31:11,534 INFO [train.py:901] (1/4) Epoch 9, batch 5150, loss[loss=0.1965, simple_loss=0.2725, pruned_loss=0.06023, over 7426.00 frames. ], tot_loss[loss=0.2492, simple_loss=0.3211, pruned_loss=0.08863, over 1621795.39 frames. ], batch size: 17, lr: 8.46e-03, grad_scale: 8.0 +2023-02-06 07:31:29,718 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.853e+02 2.410e+02 3.240e+02 3.896e+02 9.119e+02, threshold=6.481e+02, percent-clipped=3.0 +2023-02-06 07:31:32,781 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69845.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:31:46,668 INFO [train.py:901] (1/4) Epoch 9, batch 5200, loss[loss=0.2168, simple_loss=0.2939, pruned_loss=0.06987, over 7931.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.3203, pruned_loss=0.08807, over 1620736.15 frames. ], batch size: 20, lr: 8.46e-03, grad_scale: 8.0 +2023-02-06 07:31:50,870 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69871.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:06,754 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69895.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:11,378 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69901.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:12,139 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.44 vs. limit=5.0 +2023-02-06 07:32:17,775 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 07:32:20,229 INFO [train.py:901] (1/4) Epoch 9, batch 5250, loss[loss=0.2061, simple_loss=0.2854, pruned_loss=0.06339, over 7548.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.3199, pruned_loss=0.08827, over 1617152.66 frames. ], batch size: 18, lr: 8.45e-03, grad_scale: 8.0 +2023-02-06 07:32:23,774 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69920.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:23,901 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 07:32:27,862 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69926.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:38,260 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.909e+02 3.504e+02 4.160e+02 7.603e+02, threshold=7.007e+02, percent-clipped=5.0 +2023-02-06 07:32:50,718 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69958.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:55,306 INFO [train.py:901] (1/4) Epoch 9, batch 5300, loss[loss=0.2509, simple_loss=0.3309, pruned_loss=0.08551, over 8437.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3204, pruned_loss=0.08804, over 1621574.15 frames. ], batch size: 27, lr: 8.45e-03, grad_scale: 8.0 +2023-02-06 07:33:04,972 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69978.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:33:08,277 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69983.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:33:22,903 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70003.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:33:24,836 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:33:31,217 INFO [train.py:901] (1/4) Epoch 9, batch 5350, loss[loss=0.278, simple_loss=0.3508, pruned_loss=0.1026, over 8658.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3192, pruned_loss=0.08765, over 1619441.92 frames. ], batch size: 39, lr: 8.45e-03, grad_scale: 8.0 +2023-02-06 07:33:42,006 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70031.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:33:45,282 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70036.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:33:48,396 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 2.965e+02 3.484e+02 4.155e+02 9.515e+02, threshold=6.968e+02, percent-clipped=2.0 +2023-02-06 07:33:57,308 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70053.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:33:59,443 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 +2023-02-06 07:34:05,194 INFO [train.py:901] (1/4) Epoch 9, batch 5400, loss[loss=0.2493, simple_loss=0.3334, pruned_loss=0.08262, over 8509.00 frames. ], tot_loss[loss=0.2492, simple_loss=0.3207, pruned_loss=0.08886, over 1616176.76 frames. ], batch size: 26, lr: 8.45e-03, grad_scale: 8.0 +2023-02-06 07:34:14,841 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70078.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:31,586 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70101.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:40,958 INFO [train.py:901] (1/4) Epoch 9, batch 5450, loss[loss=0.2548, simple_loss=0.3165, pruned_loss=0.09652, over 8338.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3208, pruned_loss=0.08849, over 1621195.68 frames. ], batch size: 26, lr: 8.44e-03, grad_scale: 8.0 +2023-02-06 07:34:48,364 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70126.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:49,039 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70127.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:49,622 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:58,807 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.682e+02 3.191e+02 4.046e+02 1.028e+03, threshold=6.382e+02, percent-clipped=4.0 +2023-02-06 07:35:04,353 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 07:35:05,827 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70151.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:35:06,516 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70152.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:35:15,866 INFO [train.py:901] (1/4) Epoch 9, batch 5500, loss[loss=0.2158, simple_loss=0.2926, pruned_loss=0.06948, over 8490.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3223, pruned_loss=0.08981, over 1622418.25 frames. ], batch size: 28, lr: 8.44e-03, grad_scale: 16.0 +2023-02-06 07:35:50,278 INFO [train.py:901] (1/4) Epoch 9, batch 5550, loss[loss=0.3076, simple_loss=0.3527, pruned_loss=0.1313, over 7792.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3221, pruned_loss=0.08967, over 1619230.21 frames. ], batch size: 19, lr: 8.44e-03, grad_scale: 16.0 +2023-02-06 07:36:07,873 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.484e+02 3.031e+02 3.937e+02 9.276e+02, threshold=6.062e+02, percent-clipped=2.0 +2023-02-06 07:36:14,989 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 07:36:24,758 INFO [train.py:901] (1/4) Epoch 9, batch 5600, loss[loss=0.2738, simple_loss=0.3548, pruned_loss=0.09635, over 8475.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3224, pruned_loss=0.08974, over 1622807.46 frames. ], batch size: 25, lr: 8.43e-03, grad_scale: 16.0 +2023-02-06 07:36:34,075 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:36:59,380 INFO [train.py:901] (1/4) Epoch 9, batch 5650, loss[loss=0.2383, simple_loss=0.3164, pruned_loss=0.08012, over 8299.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3225, pruned_loss=0.08968, over 1620723.14 frames. ], batch size: 23, lr: 8.43e-03, grad_scale: 8.0 +2023-02-06 07:37:08,094 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 07:37:18,049 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.609e+02 3.248e+02 4.005e+02 8.106e+02, threshold=6.497e+02, percent-clipped=5.0 +2023-02-06 07:37:32,959 INFO [train.py:901] (1/4) Epoch 9, batch 5700, loss[loss=0.2339, simple_loss=0.3184, pruned_loss=0.07473, over 7665.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3217, pruned_loss=0.08966, over 1611918.12 frames. ], batch size: 19, lr: 8.43e-03, grad_scale: 8.0 +2023-02-06 07:37:38,233 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8216, 3.0753, 2.1953, 3.8429, 1.8948, 1.8365, 2.2517, 3.3233], + device='cuda:1'), covar=tensor([0.0672, 0.0779, 0.1077, 0.0286, 0.1146, 0.1569, 0.1305, 0.0789], + device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0221, 0.0262, 0.0217, 0.0222, 0.0259, 0.0266, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 07:37:55,589 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8151, 1.7340, 1.8773, 1.7802, 1.3165, 1.7952, 2.3148, 1.8648], + device='cuda:1'), covar=tensor([0.0442, 0.1112, 0.1540, 0.1241, 0.0559, 0.1315, 0.0572, 0.0548], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0159, 0.0198, 0.0163, 0.0108, 0.0167, 0.0121, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 07:38:02,682 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70407.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:38:08,022 INFO [train.py:901] (1/4) Epoch 9, batch 5750, loss[loss=0.2478, simple_loss=0.3075, pruned_loss=0.09406, over 7797.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3218, pruned_loss=0.08959, over 1614415.65 frames. ], batch size: 20, lr: 8.42e-03, grad_scale: 8.0 +2023-02-06 07:38:12,127 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 07:38:20,262 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70432.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:38:27,528 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.898e+02 3.376e+02 4.229e+02 8.555e+02, threshold=6.753e+02, percent-clipped=3.0 +2023-02-06 07:38:31,872 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3711, 2.7297, 1.8342, 3.7315, 1.6986, 1.6296, 2.2559, 3.1269], + device='cuda:1'), covar=tensor([0.0882, 0.1009, 0.1324, 0.0259, 0.1427, 0.1682, 0.1229, 0.0841], + device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0220, 0.0263, 0.0217, 0.0224, 0.0260, 0.0264, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 07:38:43,379 INFO [train.py:901] (1/4) Epoch 9, batch 5800, loss[loss=0.2755, simple_loss=0.3471, pruned_loss=0.102, over 8475.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3215, pruned_loss=0.08864, over 1616398.98 frames. ], batch size: 29, lr: 8.42e-03, grad_scale: 8.0 +2023-02-06 07:38:48,299 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:39:18,643 INFO [train.py:901] (1/4) Epoch 9, batch 5850, loss[loss=0.2572, simple_loss=0.3263, pruned_loss=0.09406, over 8079.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3189, pruned_loss=0.08721, over 1613138.44 frames. ], batch size: 21, lr: 8.42e-03, grad_scale: 8.0 +2023-02-06 07:39:37,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.477e+02 3.501e+02 4.376e+02 8.995e+02, threshold=7.001e+02, percent-clipped=4.0 +2023-02-06 07:39:48,211 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5637, 2.1070, 3.7028, 2.6301, 2.9531, 2.4128, 1.8793, 1.5006], + device='cuda:1'), covar=tensor([0.3373, 0.4003, 0.0949, 0.2371, 0.1970, 0.1926, 0.1582, 0.4386], + device='cuda:1'), in_proj_covar=tensor([0.0853, 0.0819, 0.0703, 0.0805, 0.0900, 0.0761, 0.0683, 0.0733], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 07:39:53,321 INFO [train.py:901] (1/4) Epoch 9, batch 5900, loss[loss=0.2475, simple_loss=0.3247, pruned_loss=0.08517, over 7813.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3202, pruned_loss=0.08722, over 1618082.42 frames. ], batch size: 20, lr: 8.42e-03, grad_scale: 8.0 +2023-02-06 07:40:08,009 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:40:23,561 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9539, 1.9543, 2.3280, 1.7044, 1.2293, 2.4390, 0.3877, 1.5445], + device='cuda:1'), covar=tensor([0.2363, 0.1785, 0.0472, 0.2023, 0.5186, 0.0447, 0.4055, 0.1855], + device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0160, 0.0091, 0.0210, 0.0248, 0.0096, 0.0161, 0.0158], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0001, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:40:27,233 INFO [train.py:901] (1/4) Epoch 9, batch 5950, loss[loss=0.2326, simple_loss=0.2892, pruned_loss=0.08799, over 7677.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3199, pruned_loss=0.08709, over 1619106.96 frames. ], batch size: 18, lr: 8.41e-03, grad_scale: 8.0 +2023-02-06 07:40:32,699 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:40:45,950 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.730e+02 3.193e+02 3.849e+02 7.953e+02, threshold=6.387e+02, percent-clipped=3.0 +2023-02-06 07:41:02,076 INFO [train.py:901] (1/4) Epoch 9, batch 6000, loss[loss=0.2788, simple_loss=0.3479, pruned_loss=0.1049, over 8336.00 frames. ], tot_loss[loss=0.247, simple_loss=0.32, pruned_loss=0.08702, over 1619452.98 frames. ], batch size: 25, lr: 8.41e-03, grad_scale: 8.0 +2023-02-06 07:41:02,077 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 07:41:08,245 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3266, 1.0964, 1.1670, 0.9714, 0.7347, 1.0809, 1.2010, 1.1620], + device='cuda:1'), covar=tensor([0.0468, 0.1089, 0.1473, 0.1137, 0.0508, 0.1243, 0.0566, 0.0485], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0158, 0.0198, 0.0162, 0.0108, 0.0166, 0.0121, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 07:41:14,592 INFO [train.py:935] (1/4) Epoch 9, validation: loss=0.1952, simple_loss=0.2947, pruned_loss=0.0479, over 944034.00 frames. +2023-02-06 07:41:14,593 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 07:41:33,920 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3060, 1.3891, 2.2631, 1.0774, 2.1833, 2.4280, 2.5217, 2.0510], + device='cuda:1'), covar=tensor([0.1070, 0.1181, 0.0521, 0.2020, 0.0646, 0.0410, 0.0674, 0.0822], + device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0286, 0.0250, 0.0273, 0.0265, 0.0229, 0.0312, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:41:49,702 INFO [train.py:901] (1/4) Epoch 9, batch 6050, loss[loss=0.2314, simple_loss=0.302, pruned_loss=0.08042, over 8142.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3185, pruned_loss=0.08686, over 1615516.10 frames. ], batch size: 22, lr: 8.41e-03, grad_scale: 8.0 +2023-02-06 07:41:55,399 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.59 vs. limit=5.0 +2023-02-06 07:42:04,730 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:42:07,993 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.822e+02 3.602e+02 4.348e+02 1.269e+03, threshold=7.203e+02, percent-clipped=6.0 +2023-02-06 07:42:24,360 INFO [train.py:901] (1/4) Epoch 9, batch 6100, loss[loss=0.2613, simple_loss=0.3324, pruned_loss=0.09504, over 8450.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3196, pruned_loss=0.0879, over 1613460.96 frames. ], batch size: 27, lr: 8.40e-03, grad_scale: 8.0 +2023-02-06 07:42:25,259 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3482, 1.1520, 1.4317, 1.0667, 0.7939, 1.2078, 1.1881, 0.9029], + device='cuda:1'), covar=tensor([0.0575, 0.1318, 0.1808, 0.1525, 0.0634, 0.1616, 0.0711, 0.0734], + device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0158, 0.0197, 0.0162, 0.0108, 0.0167, 0.0121, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 07:42:37,602 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2456, 1.5930, 1.5752, 0.7647, 1.6563, 1.2880, 0.2680, 1.4459], + device='cuda:1'), covar=tensor([0.0406, 0.0240, 0.0195, 0.0349, 0.0281, 0.0623, 0.0541, 0.0178], + device='cuda:1'), in_proj_covar=tensor([0.0382, 0.0304, 0.0250, 0.0363, 0.0291, 0.0455, 0.0347, 0.0331], + device='cuda:1'), out_proj_covar=tensor([1.1315e-04, 8.7876e-05, 7.2327e-05, 1.0547e-04, 8.5479e-05, 1.4409e-04, + 1.0294e-04, 9.7344e-05], device='cuda:1') +2023-02-06 07:42:42,094 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 07:42:55,120 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7115, 1.4853, 2.8073, 1.2135, 2.1170, 3.0493, 3.0226, 2.5827], + device='cuda:1'), covar=tensor([0.1011, 0.1380, 0.0403, 0.1920, 0.0783, 0.0284, 0.0518, 0.0693], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0288, 0.0251, 0.0275, 0.0267, 0.0230, 0.0313, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:42:56,500 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0783, 2.5503, 2.8965, 1.1767, 3.1823, 1.8169, 1.5906, 1.8476], + device='cuda:1'), covar=tensor([0.0448, 0.0191, 0.0165, 0.0431, 0.0212, 0.0491, 0.0465, 0.0291], + device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0300, 0.0246, 0.0360, 0.0287, 0.0449, 0.0342, 0.0327], + device='cuda:1'), out_proj_covar=tensor([1.1129e-04, 8.6525e-05, 7.1326e-05, 1.0444e-04, 8.4251e-05, 1.4203e-04, + 1.0153e-04, 9.6391e-05], device='cuda:1') +2023-02-06 07:43:00,373 INFO [train.py:901] (1/4) Epoch 9, batch 6150, loss[loss=0.2298, simple_loss=0.288, pruned_loss=0.08581, over 7532.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3192, pruned_loss=0.08772, over 1614461.25 frames. ], batch size: 18, lr: 8.40e-03, grad_scale: 8.0 +2023-02-06 07:43:18,322 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.692e+02 3.232e+02 3.879e+02 7.941e+02, threshold=6.463e+02, percent-clipped=1.0 +2023-02-06 07:43:19,251 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70843.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:43:22,705 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2268, 1.9171, 1.7561, 1.7544, 1.2457, 1.7245, 2.3300, 2.1375], + device='cuda:1'), covar=tensor([0.0410, 0.1080, 0.1684, 0.1345, 0.0575, 0.1398, 0.0589, 0.0563], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0158, 0.0198, 0.0163, 0.0108, 0.0167, 0.0121, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 07:43:33,935 INFO [train.py:901] (1/4) Epoch 9, batch 6200, loss[loss=0.2557, simple_loss=0.3221, pruned_loss=0.09463, over 7813.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3198, pruned_loss=0.08803, over 1610424.05 frames. ], batch size: 20, lr: 8.40e-03, grad_scale: 8.0 +2023-02-06 07:43:36,218 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:44:04,565 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 +2023-02-06 07:44:09,711 INFO [train.py:901] (1/4) Epoch 9, batch 6250, loss[loss=0.2945, simple_loss=0.3615, pruned_loss=0.1137, over 8102.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3192, pruned_loss=0.08778, over 1607039.77 frames. ], batch size: 23, lr: 8.40e-03, grad_scale: 8.0 +2023-02-06 07:44:28,171 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 07:44:28,454 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.792e+02 3.423e+02 4.432e+02 1.474e+03, threshold=6.847e+02, percent-clipped=7.0 +2023-02-06 07:44:42,861 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 07:44:43,938 INFO [train.py:901] (1/4) Epoch 9, batch 6300, loss[loss=0.2295, simple_loss=0.3106, pruned_loss=0.07421, over 8107.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3194, pruned_loss=0.0878, over 1602158.99 frames. ], batch size: 23, lr: 8.39e-03, grad_scale: 8.0 +2023-02-06 07:45:03,885 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:45:11,480 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9899, 2.3350, 1.6913, 3.0080, 1.3932, 1.5326, 1.9952, 2.5149], + device='cuda:1'), covar=tensor([0.0949, 0.0972, 0.1408, 0.0394, 0.1368, 0.1677, 0.1125, 0.0799], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0225, 0.0266, 0.0220, 0.0227, 0.0262, 0.0268, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 07:45:16,121 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71010.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:45:19,447 INFO [train.py:901] (1/4) Epoch 9, batch 6350, loss[loss=0.1788, simple_loss=0.2556, pruned_loss=0.05096, over 7542.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3195, pruned_loss=0.0878, over 1601584.88 frames. ], batch size: 18, lr: 8.39e-03, grad_scale: 8.0 +2023-02-06 07:45:21,648 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:45:38,849 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.814e+02 3.293e+02 4.210e+02 8.338e+02, threshold=6.585e+02, percent-clipped=5.0 +2023-02-06 07:45:42,976 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71048.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:45:54,194 INFO [train.py:901] (1/4) Epoch 9, batch 6400, loss[loss=0.2496, simple_loss=0.3099, pruned_loss=0.09462, over 8087.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3202, pruned_loss=0.08845, over 1604886.48 frames. ], batch size: 21, lr: 8.39e-03, grad_scale: 8.0 +2023-02-06 07:46:28,849 INFO [train.py:901] (1/4) Epoch 9, batch 6450, loss[loss=0.249, simple_loss=0.3207, pruned_loss=0.08865, over 8369.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3212, pruned_loss=0.08908, over 1609609.63 frames. ], batch size: 24, lr: 8.38e-03, grad_scale: 4.0 +2023-02-06 07:46:48,393 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.655e+02 3.350e+02 4.272e+02 1.011e+03, threshold=6.701e+02, percent-clipped=3.0 +2023-02-06 07:47:03,601 INFO [train.py:901] (1/4) Epoch 9, batch 6500, loss[loss=0.2818, simple_loss=0.3535, pruned_loss=0.1051, over 8249.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3204, pruned_loss=0.08845, over 1611433.83 frames. ], batch size: 24, lr: 8.38e-03, grad_scale: 4.0 +2023-02-06 07:47:37,717 INFO [train.py:901] (1/4) Epoch 9, batch 6550, loss[loss=0.2622, simple_loss=0.3443, pruned_loss=0.09012, over 8325.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3197, pruned_loss=0.08787, over 1604811.45 frames. ], batch size: 25, lr: 8.38e-03, grad_scale: 4.0 +2023-02-06 07:47:50,050 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 07:47:58,022 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.500e+02 3.444e+02 4.178e+02 7.414e+02, threshold=6.887e+02, percent-clipped=1.0 +2023-02-06 07:48:10,532 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 07:48:13,164 INFO [train.py:901] (1/4) Epoch 9, batch 6600, loss[loss=0.2016, simple_loss=0.2782, pruned_loss=0.0625, over 7925.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3191, pruned_loss=0.08778, over 1607437.70 frames. ], batch size: 20, lr: 8.37e-03, grad_scale: 4.0 +2023-02-06 07:48:22,791 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7745, 1.5215, 3.3660, 1.3428, 2.3204, 3.6066, 3.6149, 3.0533], + device='cuda:1'), covar=tensor([0.1081, 0.1501, 0.0298, 0.2062, 0.0895, 0.0237, 0.0342, 0.0611], + device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0285, 0.0248, 0.0279, 0.0263, 0.0229, 0.0308, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:48:32,873 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.48 vs. limit=5.0 +2023-02-06 07:48:47,048 INFO [train.py:901] (1/4) Epoch 9, batch 6650, loss[loss=0.2434, simple_loss=0.3153, pruned_loss=0.08576, over 8099.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3186, pruned_loss=0.08694, over 1610194.93 frames. ], batch size: 23, lr: 8.37e-03, grad_scale: 4.0 +2023-02-06 07:49:05,719 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.641e+02 3.214e+02 4.234e+02 1.005e+03, threshold=6.427e+02, percent-clipped=4.0 +2023-02-06 07:49:12,673 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3978, 4.3076, 3.8725, 2.3615, 3.7893, 3.8850, 4.0654, 3.4867], + device='cuda:1'), covar=tensor([0.0753, 0.0631, 0.1032, 0.4344, 0.0750, 0.0905, 0.1231, 0.0975], + device='cuda:1'), in_proj_covar=tensor([0.0451, 0.0359, 0.0366, 0.0470, 0.0372, 0.0353, 0.0364, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 07:49:13,931 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71354.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:49:21,687 INFO [train.py:901] (1/4) Epoch 9, batch 6700, loss[loss=0.2568, simple_loss=0.3305, pruned_loss=0.09154, over 8338.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3194, pruned_loss=0.08761, over 1609313.75 frames. ], batch size: 26, lr: 8.37e-03, grad_scale: 4.0 +2023-02-06 07:49:41,043 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:49:56,528 INFO [train.py:901] (1/4) Epoch 9, batch 6750, loss[loss=0.2211, simple_loss=0.2869, pruned_loss=0.07769, over 7527.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3189, pruned_loss=0.08716, over 1608552.07 frames. ], batch size: 18, lr: 8.37e-03, grad_scale: 4.0 +2023-02-06 07:50:15,374 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 3.032e+02 3.821e+02 4.704e+02 1.129e+03, threshold=7.641e+02, percent-clipped=7.0 +2023-02-06 07:50:23,372 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 07:50:30,405 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2969, 1.5173, 2.2367, 1.2052, 1.5319, 1.5410, 1.4453, 1.4187], + device='cuda:1'), covar=tensor([0.1718, 0.1933, 0.0677, 0.3388, 0.1481, 0.2713, 0.1652, 0.1803], + device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0500, 0.0529, 0.0575, 0.0611, 0.0548, 0.0469, 0.0612], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 07:50:30,842 INFO [train.py:901] (1/4) Epoch 9, batch 6800, loss[loss=0.2716, simple_loss=0.3371, pruned_loss=0.103, over 8630.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3188, pruned_loss=0.08716, over 1609913.20 frames. ], batch size: 31, lr: 8.36e-03, grad_scale: 8.0 +2023-02-06 07:50:33,613 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71469.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:51:01,250 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:51:06,464 INFO [train.py:901] (1/4) Epoch 9, batch 6850, loss[loss=0.2497, simple_loss=0.3253, pruned_loss=0.0871, over 8475.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3197, pruned_loss=0.08825, over 1611101.04 frames. ], batch size: 25, lr: 8.36e-03, grad_scale: 8.0 +2023-02-06 07:51:14,494 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 07:51:25,240 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.628e+02 3.217e+02 4.054e+02 6.964e+02, threshold=6.433e+02, percent-clipped=0.0 +2023-02-06 07:51:40,035 INFO [train.py:901] (1/4) Epoch 9, batch 6900, loss[loss=0.2606, simple_loss=0.3436, pruned_loss=0.08877, over 8028.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3196, pruned_loss=0.08771, over 1613722.57 frames. ], batch size: 22, lr: 8.36e-03, grad_scale: 8.0 +2023-02-06 07:51:48,539 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2272, 1.4168, 3.0927, 1.1621, 2.1745, 3.3852, 3.6340, 2.4697], + device='cuda:1'), covar=tensor([0.1009, 0.1793, 0.0512, 0.2590, 0.1213, 0.0410, 0.0549, 0.1245], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0282, 0.0249, 0.0278, 0.0261, 0.0226, 0.0305, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:52:15,382 INFO [train.py:901] (1/4) Epoch 9, batch 6950, loss[loss=0.2262, simple_loss=0.2962, pruned_loss=0.07813, over 7661.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3205, pruned_loss=0.08846, over 1616175.53 frames. ], batch size: 19, lr: 8.35e-03, grad_scale: 8.0 +2023-02-06 07:52:23,475 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 07:52:29,644 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:52:35,519 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.622e+02 3.284e+02 3.978e+02 8.428e+02, threshold=6.567e+02, percent-clipped=2.0 +2023-02-06 07:52:50,437 INFO [train.py:901] (1/4) Epoch 9, batch 7000, loss[loss=0.2566, simple_loss=0.3298, pruned_loss=0.0917, over 8243.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3204, pruned_loss=0.08767, over 1619996.64 frames. ], batch size: 24, lr: 8.35e-03, grad_scale: 8.0 +2023-02-06 07:53:19,913 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 07:53:21,191 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-02-06 07:53:24,874 INFO [train.py:901] (1/4) Epoch 9, batch 7050, loss[loss=0.2534, simple_loss=0.3273, pruned_loss=0.08973, over 8581.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3202, pruned_loss=0.08742, over 1617098.80 frames. ], batch size: 39, lr: 8.35e-03, grad_scale: 8.0 +2023-02-06 07:53:32,436 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71725.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:53:45,154 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.687e+02 2.886e+02 3.338e+02 4.007e+02 6.250e+02, threshold=6.676e+02, percent-clipped=0.0 +2023-02-06 07:53:48,781 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9119, 1.5895, 6.0263, 2.1329, 5.2433, 5.0324, 5.6399, 5.3887], + device='cuda:1'), covar=tensor([0.0522, 0.4609, 0.0413, 0.3231, 0.1189, 0.0849, 0.0448, 0.0533], + device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0552, 0.0553, 0.0513, 0.0582, 0.0495, 0.0485, 0.0545], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 07:53:50,707 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:53:59,284 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71763.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:54:00,448 INFO [train.py:901] (1/4) Epoch 9, batch 7100, loss[loss=0.2609, simple_loss=0.3286, pruned_loss=0.09659, over 8287.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3192, pruned_loss=0.08651, over 1616165.55 frames. ], batch size: 23, lr: 8.35e-03, grad_scale: 8.0 +2023-02-06 07:54:08,143 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 07:54:16,149 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71788.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:54:34,523 INFO [train.py:901] (1/4) Epoch 9, batch 7150, loss[loss=0.2372, simple_loss=0.3207, pruned_loss=0.07686, over 8178.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3186, pruned_loss=0.08609, over 1616015.13 frames. ], batch size: 23, lr: 8.34e-03, grad_scale: 8.0 +2023-02-06 07:54:54,748 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.560e+02 3.246e+02 4.043e+02 1.359e+03, threshold=6.493e+02, percent-clipped=7.0 +2023-02-06 07:55:10,759 INFO [train.py:901] (1/4) Epoch 9, batch 7200, loss[loss=0.2384, simple_loss=0.3043, pruned_loss=0.08624, over 7647.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3196, pruned_loss=0.08696, over 1614825.41 frames. ], batch size: 19, lr: 8.34e-03, grad_scale: 8.0 +2023-02-06 07:55:37,710 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 07:55:40,054 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4222, 1.4783, 1.6622, 1.3209, 0.8938, 1.6569, 0.1940, 1.1318], + device='cuda:1'), covar=tensor([0.2664, 0.1775, 0.0616, 0.1662, 0.4503, 0.0599, 0.3521, 0.1879], + device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0157, 0.0092, 0.0205, 0.0244, 0.0095, 0.0154, 0.0155], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 07:55:41,084 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.85 vs. limit=5.0 +2023-02-06 07:55:43,959 INFO [train.py:901] (1/4) Epoch 9, batch 7250, loss[loss=0.1736, simple_loss=0.2547, pruned_loss=0.04625, over 7314.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3199, pruned_loss=0.0871, over 1617743.00 frames. ], batch size: 16, lr: 8.34e-03, grad_scale: 8.0 +2023-02-06 07:55:58,954 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71937.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:56:02,879 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.597e+02 3.277e+02 3.984e+02 9.565e+02, threshold=6.554e+02, percent-clipped=6.0 +2023-02-06 07:56:19,573 INFO [train.py:901] (1/4) Epoch 9, batch 7300, loss[loss=0.3104, simple_loss=0.36, pruned_loss=0.1304, over 6757.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3195, pruned_loss=0.0868, over 1619012.33 frames. ], batch size: 71, lr: 8.33e-03, grad_scale: 8.0 +2023-02-06 07:56:28,820 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71978.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:56:54,487 INFO [train.py:901] (1/4) Epoch 9, batch 7350, loss[loss=0.258, simple_loss=0.3318, pruned_loss=0.09211, over 8359.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3206, pruned_loss=0.08743, over 1623717.33 frames. ], batch size: 24, lr: 8.33e-03, grad_scale: 8.0 +2023-02-06 07:57:02,503 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 07:57:04,267 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.87 vs. limit=5.0 +2023-02-06 07:57:13,529 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.007e+02 2.925e+02 3.749e+02 4.804e+02 1.068e+03, threshold=7.499e+02, percent-clipped=9.0 +2023-02-06 07:57:22,479 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 07:57:30,035 INFO [train.py:901] (1/4) Epoch 9, batch 7400, loss[loss=0.3034, simple_loss=0.3651, pruned_loss=0.1208, over 8546.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.32, pruned_loss=0.08659, over 1625126.08 frames. ], batch size: 49, lr: 8.33e-03, grad_scale: 8.0 +2023-02-06 07:57:40,384 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4709, 1.7882, 2.9040, 1.3194, 1.9899, 1.9004, 1.5411, 1.7925], + device='cuda:1'), covar=tensor([0.1597, 0.1886, 0.0563, 0.3459, 0.1455, 0.2499, 0.1619, 0.1902], + device='cuda:1'), in_proj_covar=tensor([0.0481, 0.0494, 0.0530, 0.0569, 0.0608, 0.0539, 0.0463, 0.0609], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 07:57:50,406 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72093.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:58:03,870 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 07:58:05,139 INFO [train.py:901] (1/4) Epoch 9, batch 7450, loss[loss=0.2813, simple_loss=0.3573, pruned_loss=0.1027, over 8472.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3203, pruned_loss=0.08714, over 1625218.82 frames. ], batch size: 25, lr: 8.33e-03, grad_scale: 8.0 +2023-02-06 07:58:23,972 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.502e+02 2.410e+02 3.229e+02 3.860e+02 9.903e+02, threshold=6.459e+02, percent-clipped=1.0 +2023-02-06 07:58:26,826 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72147.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:58:38,697 INFO [train.py:901] (1/4) Epoch 9, batch 7500, loss[loss=0.2082, simple_loss=0.2807, pruned_loss=0.0679, over 7789.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3192, pruned_loss=0.0868, over 1623941.07 frames. ], batch size: 19, lr: 8.32e-03, grad_scale: 8.0 +2023-02-06 07:59:15,024 INFO [train.py:901] (1/4) Epoch 9, batch 7550, loss[loss=0.2534, simple_loss=0.3221, pruned_loss=0.09237, over 8071.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3192, pruned_loss=0.08682, over 1621418.65 frames. ], batch size: 21, lr: 8.32e-03, grad_scale: 8.0 +2023-02-06 07:59:33,698 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 2.819e+02 3.433e+02 4.309e+02 8.597e+02, threshold=6.865e+02, percent-clipped=4.0 +2023-02-06 07:59:48,157 INFO [train.py:901] (1/4) Epoch 9, batch 7600, loss[loss=0.2822, simple_loss=0.3491, pruned_loss=0.1077, over 8360.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3206, pruned_loss=0.08782, over 1617889.25 frames. ], batch size: 24, lr: 8.32e-03, grad_scale: 8.0 +2023-02-06 07:59:48,738 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 07:59:54,369 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 +2023-02-06 07:59:58,697 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72281.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:00:22,667 INFO [train.py:901] (1/4) Epoch 9, batch 7650, loss[loss=0.2355, simple_loss=0.3168, pruned_loss=0.07709, over 8189.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3205, pruned_loss=0.08783, over 1619011.73 frames. ], batch size: 23, lr: 8.31e-03, grad_scale: 8.0 +2023-02-06 08:00:42,776 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.706e+02 3.178e+02 3.983e+02 6.818e+02, threshold=6.357e+02, percent-clipped=0.0 +2023-02-06 08:00:43,593 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72344.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:00:47,001 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:00:57,587 INFO [train.py:901] (1/4) Epoch 9, batch 7700, loss[loss=0.3303, simple_loss=0.3769, pruned_loss=0.1419, over 6851.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.319, pruned_loss=0.08705, over 1616520.53 frames. ], batch size: 71, lr: 8.31e-03, grad_scale: 8.0 +2023-02-06 08:01:03,922 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72374.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:01:09,242 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9008, 1.5682, 3.1407, 1.3176, 2.2954, 3.2642, 3.4047, 2.8099], + device='cuda:1'), covar=tensor([0.1002, 0.1486, 0.0339, 0.2098, 0.0839, 0.0333, 0.0541, 0.0717], + device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0293, 0.0252, 0.0283, 0.0268, 0.0234, 0.0316, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 08:01:09,787 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 08:01:18,755 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72396.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:01:32,214 INFO [train.py:901] (1/4) Epoch 9, batch 7750, loss[loss=0.2422, simple_loss=0.3244, pruned_loss=0.07999, over 8468.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.319, pruned_loss=0.08658, over 1617841.16 frames. ], batch size: 29, lr: 8.31e-03, grad_scale: 8.0 +2023-02-06 08:01:42,786 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 08:01:53,031 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.670e+02 3.267e+02 4.054e+02 1.108e+03, threshold=6.534e+02, percent-clipped=5.0 +2023-02-06 08:02:08,196 INFO [train.py:901] (1/4) Epoch 9, batch 7800, loss[loss=0.2, simple_loss=0.2848, pruned_loss=0.05763, over 8240.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3189, pruned_loss=0.08648, over 1615535.97 frames. ], batch size: 22, lr: 8.31e-03, grad_scale: 8.0 +2023-02-06 08:02:12,663 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 +2023-02-06 08:02:22,373 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1850, 1.8750, 2.7996, 2.1966, 2.4388, 2.0508, 1.6140, 1.1950], + device='cuda:1'), covar=tensor([0.3706, 0.3715, 0.0958, 0.2285, 0.1808, 0.1997, 0.1641, 0.3848], + device='cuda:1'), in_proj_covar=tensor([0.0861, 0.0827, 0.0699, 0.0815, 0.0902, 0.0763, 0.0690, 0.0745], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:02:25,493 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72491.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:02:41,270 INFO [train.py:901] (1/4) Epoch 9, batch 7850, loss[loss=0.2375, simple_loss=0.3072, pruned_loss=0.08387, over 7922.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3178, pruned_loss=0.08547, over 1614389.24 frames. ], batch size: 20, lr: 8.30e-03, grad_scale: 8.0 +2023-02-06 08:02:59,530 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.530e+02 3.201e+02 3.890e+02 8.475e+02, threshold=6.403e+02, percent-clipped=6.0 +2023-02-06 08:03:14,039 INFO [train.py:901] (1/4) Epoch 9, batch 7900, loss[loss=0.241, simple_loss=0.3173, pruned_loss=0.08239, over 7809.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3173, pruned_loss=0.08575, over 1610438.86 frames. ], batch size: 20, lr: 8.30e-03, grad_scale: 8.0 +2023-02-06 08:03:14,273 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4693, 1.8376, 3.4510, 1.2181, 2.3881, 1.8364, 1.4729, 2.2502], + device='cuda:1'), covar=tensor([0.1710, 0.2301, 0.0788, 0.3919, 0.1754, 0.2902, 0.1889, 0.2482], + device='cuda:1'), in_proj_covar=tensor([0.0477, 0.0493, 0.0527, 0.0568, 0.0605, 0.0539, 0.0464, 0.0609], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 08:03:31,539 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7568, 3.7401, 3.4226, 1.5679, 3.3464, 3.3034, 3.4876, 3.1013], + device='cuda:1'), covar=tensor([0.0888, 0.0667, 0.1028, 0.5242, 0.0858, 0.1016, 0.1246, 0.0937], + device='cuda:1'), in_proj_covar=tensor([0.0449, 0.0353, 0.0367, 0.0467, 0.0368, 0.0349, 0.0360, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 08:03:39,489 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0556, 1.4467, 3.3555, 1.4110, 2.1675, 3.6949, 3.7372, 3.1059], + device='cuda:1'), covar=tensor([0.0966, 0.1500, 0.0373, 0.2098, 0.1098, 0.0272, 0.0490, 0.0706], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0290, 0.0252, 0.0281, 0.0264, 0.0231, 0.0314, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 08:03:41,420 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72606.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:03:47,073 INFO [train.py:901] (1/4) Epoch 9, batch 7950, loss[loss=0.2318, simple_loss=0.3074, pruned_loss=0.07805, over 7911.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3179, pruned_loss=0.08576, over 1615918.52 frames. ], batch size: 20, lr: 8.30e-03, grad_scale: 8.0 +2023-02-06 08:04:05,347 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.498e+02 3.176e+02 4.184e+02 8.861e+02, threshold=6.353e+02, percent-clipped=6.0 +2023-02-06 08:04:11,439 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72652.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:04:19,757 INFO [train.py:901] (1/4) Epoch 9, batch 8000, loss[loss=0.2416, simple_loss=0.3227, pruned_loss=0.08021, over 8369.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3178, pruned_loss=0.08595, over 1615331.31 frames. ], batch size: 49, lr: 8.29e-03, grad_scale: 8.0 +2023-02-06 08:04:27,790 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72677.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:04:34,986 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72688.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:04:52,985 INFO [train.py:901] (1/4) Epoch 9, batch 8050, loss[loss=0.2924, simple_loss=0.349, pruned_loss=0.1179, over 6895.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3167, pruned_loss=0.08654, over 1590111.34 frames. ], batch size: 71, lr: 8.29e-03, grad_scale: 8.0 +2023-02-06 08:05:11,531 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 2.635e+02 3.102e+02 3.711e+02 7.462e+02, threshold=6.205e+02, percent-clipped=1.0 +2023-02-06 08:05:25,695 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 08:05:31,303 INFO [train.py:901] (1/4) Epoch 10, batch 0, loss[loss=0.22, simple_loss=0.2946, pruned_loss=0.07268, over 8246.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2946, pruned_loss=0.07268, over 8246.00 frames. ], batch size: 22, lr: 7.88e-03, grad_scale: 8.0 +2023-02-06 08:05:31,303 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 08:05:43,259 INFO [train.py:935] (1/4) Epoch 10, validation: loss=0.1954, simple_loss=0.295, pruned_loss=0.0479, over 944034.00 frames. +2023-02-06 08:05:43,259 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 08:05:57,157 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 08:06:07,500 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 08:06:17,953 INFO [train.py:901] (1/4) Epoch 10, batch 50, loss[loss=0.2902, simple_loss=0.3449, pruned_loss=0.1178, over 7925.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3192, pruned_loss=0.08828, over 366393.45 frames. ], batch size: 20, lr: 7.88e-03, grad_scale: 8.0 +2023-02-06 08:06:21,761 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72803.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:06:31,232 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 08:06:49,408 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.716e+02 3.124e+02 3.887e+02 7.160e+02, threshold=6.248e+02, percent-clipped=5.0 +2023-02-06 08:06:52,302 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 08:06:52,960 INFO [train.py:901] (1/4) Epoch 10, batch 100, loss[loss=0.247, simple_loss=0.3212, pruned_loss=0.08642, over 8239.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3206, pruned_loss=0.08807, over 645289.54 frames. ], batch size: 24, lr: 7.88e-03, grad_scale: 8.0 +2023-02-06 08:07:03,813 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72862.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:07:22,311 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72887.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:07:30,273 INFO [train.py:901] (1/4) Epoch 10, batch 150, loss[loss=0.2419, simple_loss=0.3188, pruned_loss=0.08247, over 8140.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.3189, pruned_loss=0.08682, over 858361.05 frames. ], batch size: 22, lr: 7.87e-03, grad_scale: 8.0 +2023-02-06 08:07:33,249 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2499, 1.8797, 3.0509, 2.4117, 2.7002, 2.0419, 1.6016, 1.3971], + device='cuda:1'), covar=tensor([0.3538, 0.3775, 0.0888, 0.2267, 0.1801, 0.2096, 0.1744, 0.3730], + device='cuda:1'), in_proj_covar=tensor([0.0865, 0.0828, 0.0700, 0.0817, 0.0905, 0.0763, 0.0689, 0.0744], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:07:36,683 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9315, 1.5806, 3.1608, 1.3776, 2.2149, 3.5330, 3.4893, 3.0086], + device='cuda:1'), covar=tensor([0.1125, 0.1564, 0.0421, 0.2221, 0.1010, 0.0270, 0.0536, 0.0666], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0289, 0.0251, 0.0279, 0.0264, 0.0230, 0.0312, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 08:07:40,079 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72912.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:08:01,152 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.670e+02 3.307e+02 4.288e+02 9.841e+02, threshold=6.614e+02, percent-clipped=3.0 +2023-02-06 08:08:04,550 INFO [train.py:901] (1/4) Epoch 10, batch 200, loss[loss=0.2825, simple_loss=0.3538, pruned_loss=0.1056, over 8775.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3185, pruned_loss=0.08667, over 1024182.55 frames. ], batch size: 34, lr: 7.87e-03, grad_scale: 8.0 +2023-02-06 08:08:29,432 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72982.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:08:41,007 INFO [train.py:901] (1/4) Epoch 10, batch 250, loss[loss=0.2884, simple_loss=0.3465, pruned_loss=0.1151, over 8248.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3183, pruned_loss=0.08607, over 1158039.68 frames. ], batch size: 22, lr: 7.87e-03, grad_scale: 8.0 +2023-02-06 08:08:47,831 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 08:08:56,851 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 08:09:02,436 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7115, 1.4168, 1.4906, 1.3325, 0.8998, 1.2967, 1.5014, 1.2216], + device='cuda:1'), covar=tensor([0.0571, 0.1216, 0.1715, 0.1349, 0.0584, 0.1521, 0.0699, 0.0649], + device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0155, 0.0196, 0.0160, 0.0107, 0.0167, 0.0120, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 08:09:07,478 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 08:09:12,563 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.688e+02 3.158e+02 3.760e+02 5.735e+02, threshold=6.316e+02, percent-clipped=0.0 +2023-02-06 08:09:16,038 INFO [train.py:901] (1/4) Epoch 10, batch 300, loss[loss=0.2181, simple_loss=0.2811, pruned_loss=0.07754, over 8084.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3189, pruned_loss=0.08623, over 1261071.37 frames. ], batch size: 21, lr: 7.87e-03, grad_scale: 8.0 +2023-02-06 08:09:23,775 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:09:40,902 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73084.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:09:51,612 INFO [train.py:901] (1/4) Epoch 10, batch 350, loss[loss=0.2263, simple_loss=0.3046, pruned_loss=0.07395, over 8101.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3206, pruned_loss=0.08698, over 1344765.89 frames. ], batch size: 23, lr: 7.86e-03, grad_scale: 16.0 +2023-02-06 08:09:55,111 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5642, 2.0000, 3.2447, 1.3059, 2.3411, 1.9516, 1.5724, 2.1558], + device='cuda:1'), covar=tensor([0.1600, 0.1995, 0.0766, 0.3673, 0.1489, 0.2576, 0.1726, 0.2164], + device='cuda:1'), in_proj_covar=tensor([0.0481, 0.0495, 0.0534, 0.0567, 0.0607, 0.0540, 0.0466, 0.0605], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 08:10:23,499 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 2.632e+02 3.058e+02 3.924e+02 7.931e+02, threshold=6.116e+02, percent-clipped=5.0 +2023-02-06 08:10:26,909 INFO [train.py:901] (1/4) Epoch 10, batch 400, loss[loss=0.301, simple_loss=0.3682, pruned_loss=0.117, over 8462.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3218, pruned_loss=0.0876, over 1407336.87 frames. ], batch size: 27, lr: 7.86e-03, grad_scale: 16.0 +2023-02-06 08:10:49,313 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5532, 2.7657, 1.8246, 2.1699, 2.1426, 1.5116, 1.9788, 2.1363], + device='cuda:1'), covar=tensor([0.1439, 0.0339, 0.1108, 0.0622, 0.0616, 0.1348, 0.0971, 0.0983], + device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0231, 0.0310, 0.0297, 0.0300, 0.0318, 0.0335, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 08:11:01,347 INFO [train.py:901] (1/4) Epoch 10, batch 450, loss[loss=0.1895, simple_loss=0.2525, pruned_loss=0.06323, over 7699.00 frames. ], tot_loss[loss=0.249, simple_loss=0.322, pruned_loss=0.08798, over 1452227.97 frames. ], batch size: 18, lr: 7.86e-03, grad_scale: 16.0 +2023-02-06 08:11:33,881 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 2.629e+02 3.140e+02 3.877e+02 8.143e+02, threshold=6.279e+02, percent-clipped=4.0 +2023-02-06 08:11:35,981 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.3639, 1.3668, 5.7374, 2.2170, 4.5285, 4.6563, 5.3490, 5.2444], + device='cuda:1'), covar=tensor([0.0955, 0.7136, 0.0743, 0.4437, 0.2126, 0.1569, 0.0836, 0.0707], + device='cuda:1'), in_proj_covar=tensor([0.0452, 0.0566, 0.0562, 0.0524, 0.0594, 0.0510, 0.0496, 0.0559], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:11:37,159 INFO [train.py:901] (1/4) Epoch 10, batch 500, loss[loss=0.2719, simple_loss=0.3404, pruned_loss=0.1017, over 8337.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3209, pruned_loss=0.0877, over 1489094.03 frames. ], batch size: 26, lr: 7.86e-03, grad_scale: 16.0 +2023-02-06 08:11:42,548 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73256.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:12:10,501 INFO [train.py:901] (1/4) Epoch 10, batch 550, loss[loss=0.2148, simple_loss=0.2935, pruned_loss=0.0681, over 7945.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3199, pruned_loss=0.0869, over 1518608.27 frames. ], batch size: 20, lr: 7.85e-03, grad_scale: 16.0 +2023-02-06 08:12:18,663 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7399, 1.5236, 3.0544, 1.4343, 2.1328, 3.2604, 3.3146, 2.8060], + device='cuda:1'), covar=tensor([0.1081, 0.1457, 0.0428, 0.1917, 0.1015, 0.0286, 0.0554, 0.0638], + device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0290, 0.0252, 0.0282, 0.0265, 0.0231, 0.0314, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 08:12:19,366 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73311.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:12:19,486 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5872, 1.9272, 3.0425, 2.3866, 2.7311, 2.2159, 1.8063, 1.3797], + device='cuda:1'), covar=tensor([0.2963, 0.3513, 0.0842, 0.2219, 0.1603, 0.1866, 0.1550, 0.3468], + device='cuda:1'), in_proj_covar=tensor([0.0860, 0.0831, 0.0701, 0.0817, 0.0907, 0.0763, 0.0690, 0.0744], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:12:23,932 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1351, 1.4565, 1.4836, 1.3115, 1.1145, 1.3680, 1.7093, 1.6311], + device='cuda:1'), covar=tensor([0.0505, 0.1240, 0.1828, 0.1437, 0.0594, 0.1539, 0.0707, 0.0594], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0157, 0.0197, 0.0161, 0.0108, 0.0168, 0.0120, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 08:12:29,157 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73326.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:12:41,607 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.475e+02 3.100e+02 3.629e+02 1.040e+03, threshold=6.201e+02, percent-clipped=3.0 +2023-02-06 08:12:44,825 INFO [train.py:901] (1/4) Epoch 10, batch 600, loss[loss=0.2563, simple_loss=0.3165, pruned_loss=0.09808, over 7805.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3173, pruned_loss=0.08585, over 1536988.66 frames. ], batch size: 20, lr: 7.85e-03, grad_scale: 16.0 +2023-02-06 08:12:56,198 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 08:13:01,726 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73371.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:13:20,007 INFO [train.py:901] (1/4) Epoch 10, batch 650, loss[loss=0.2307, simple_loss=0.3085, pruned_loss=0.07651, over 8105.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3174, pruned_loss=0.08608, over 1550066.82 frames. ], batch size: 23, lr: 7.85e-03, grad_scale: 16.0 +2023-02-06 08:13:50,088 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73441.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:13:51,167 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.469e+02 3.040e+02 3.840e+02 6.530e+02, threshold=6.081e+02, percent-clipped=1.0 +2023-02-06 08:13:54,587 INFO [train.py:901] (1/4) Epoch 10, batch 700, loss[loss=0.2401, simple_loss=0.3216, pruned_loss=0.0793, over 8546.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3168, pruned_loss=0.08546, over 1562854.18 frames. ], batch size: 28, lr: 7.84e-03, grad_scale: 16.0 +2023-02-06 08:14:31,475 INFO [train.py:901] (1/4) Epoch 10, batch 750, loss[loss=0.2503, simple_loss=0.327, pruned_loss=0.08678, over 8323.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3171, pruned_loss=0.08554, over 1574533.55 frames. ], batch size: 25, lr: 7.84e-03, grad_scale: 16.0 +2023-02-06 08:14:45,797 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 08:14:54,757 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 08:15:02,253 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.756e+02 3.307e+02 3.958e+02 8.111e+02, threshold=6.615e+02, percent-clipped=6.0 +2023-02-06 08:15:02,470 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0108, 2.4234, 1.9284, 2.8933, 1.4446, 1.6119, 1.7595, 2.4711], + device='cuda:1'), covar=tensor([0.0836, 0.0892, 0.1029, 0.0397, 0.1249, 0.1606, 0.1287, 0.0784], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0221, 0.0264, 0.0220, 0.0223, 0.0260, 0.0267, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 08:15:05,716 INFO [train.py:901] (1/4) Epoch 10, batch 800, loss[loss=0.2456, simple_loss=0.3227, pruned_loss=0.08425, over 8464.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3174, pruned_loss=0.08545, over 1586140.99 frames. ], batch size: 27, lr: 7.84e-03, grad_scale: 16.0 +2023-02-06 08:15:17,220 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7375, 1.3494, 3.9537, 1.3770, 3.4775, 3.3801, 3.5679, 3.4788], + device='cuda:1'), covar=tensor([0.0651, 0.3858, 0.0613, 0.3313, 0.1350, 0.0942, 0.0612, 0.0703], + device='cuda:1'), in_proj_covar=tensor([0.0444, 0.0556, 0.0557, 0.0510, 0.0585, 0.0496, 0.0488, 0.0552], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:15:41,986 INFO [train.py:901] (1/4) Epoch 10, batch 850, loss[loss=0.262, simple_loss=0.3351, pruned_loss=0.0944, over 8735.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3177, pruned_loss=0.08583, over 1597564.35 frames. ], batch size: 40, lr: 7.84e-03, grad_scale: 16.0 +2023-02-06 08:15:49,887 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:16:03,000 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73627.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:16:13,765 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.847e+02 3.470e+02 4.482e+02 1.720e+03, threshold=6.940e+02, percent-clipped=10.0 +2023-02-06 08:16:17,262 INFO [train.py:901] (1/4) Epoch 10, batch 900, loss[loss=0.2238, simple_loss=0.3083, pruned_loss=0.06965, over 8095.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3174, pruned_loss=0.08563, over 1594883.58 frames. ], batch size: 23, lr: 7.83e-03, grad_scale: 16.0 +2023-02-06 08:16:20,225 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73652.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:16:22,222 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73655.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:16:52,136 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73697.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:16:52,447 INFO [train.py:901] (1/4) Epoch 10, batch 950, loss[loss=0.3178, simple_loss=0.3446, pruned_loss=0.1455, over 7682.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.318, pruned_loss=0.08624, over 1601369.38 frames. ], batch size: 18, lr: 7.83e-03, grad_scale: 8.0 +2023-02-06 08:17:06,245 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.5677, 1.6429, 3.6618, 1.9830, 3.3209, 3.1307, 3.4060, 3.3159], + device='cuda:1'), covar=tensor([0.0534, 0.3270, 0.0722, 0.2696, 0.0948, 0.0809, 0.0484, 0.0558], + device='cuda:1'), in_proj_covar=tensor([0.0443, 0.0553, 0.0555, 0.0507, 0.0585, 0.0496, 0.0486, 0.0549], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:17:10,352 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73722.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:17:18,334 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 08:17:24,921 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.751e+02 3.323e+02 4.211e+02 1.163e+03, threshold=6.645e+02, percent-clipped=9.0 +2023-02-06 08:17:27,463 INFO [train.py:901] (1/4) Epoch 10, batch 1000, loss[loss=0.2233, simple_loss=0.3015, pruned_loss=0.07252, over 7926.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3172, pruned_loss=0.08627, over 1599451.77 frames. ], batch size: 20, lr: 7.83e-03, grad_scale: 8.0 +2023-02-06 08:17:28,906 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:17:35,435 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3454, 1.3669, 4.4991, 1.7696, 3.9708, 3.7486, 4.0862, 3.9171], + device='cuda:1'), covar=tensor([0.0496, 0.4139, 0.0470, 0.3120, 0.1083, 0.0807, 0.0511, 0.0586], + device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0563, 0.0564, 0.0511, 0.0591, 0.0504, 0.0492, 0.0557], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:17:36,340 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.17 vs. limit=5.0 +2023-02-06 08:17:42,252 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73770.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:17:50,830 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 08:18:00,668 INFO [train.py:901] (1/4) Epoch 10, batch 1050, loss[loss=0.2932, simple_loss=0.3507, pruned_loss=0.1179, over 8423.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3174, pruned_loss=0.08651, over 1605832.51 frames. ], batch size: 48, lr: 7.83e-03, grad_scale: 8.0 +2023-02-06 08:18:01,396 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 08:18:34,181 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.634e+02 3.058e+02 3.903e+02 1.179e+03, threshold=6.116e+02, percent-clipped=2.0 +2023-02-06 08:18:36,807 INFO [train.py:901] (1/4) Epoch 10, batch 1100, loss[loss=0.2997, simple_loss=0.3666, pruned_loss=0.1164, over 8606.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3167, pruned_loss=0.08565, over 1611832.57 frames. ], batch size: 31, lr: 7.82e-03, grad_scale: 8.0 +2023-02-06 08:19:09,836 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 08:19:10,510 INFO [train.py:901] (1/4) Epoch 10, batch 1150, loss[loss=0.2463, simple_loss=0.3207, pruned_loss=0.08598, over 8139.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3171, pruned_loss=0.08522, over 1614600.99 frames. ], batch size: 22, lr: 7.82e-03, grad_scale: 8.0 +2023-02-06 08:19:42,447 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.356e+02 2.791e+02 3.726e+02 1.227e+03, threshold=5.583e+02, percent-clipped=4.0 +2023-02-06 08:19:45,139 INFO [train.py:901] (1/4) Epoch 10, batch 1200, loss[loss=0.2298, simple_loss=0.3054, pruned_loss=0.07709, over 8262.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3154, pruned_loss=0.08355, over 1612730.31 frames. ], batch size: 22, lr: 7.82e-03, grad_scale: 8.0 +2023-02-06 08:19:48,561 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73952.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:20:20,063 INFO [train.py:901] (1/4) Epoch 10, batch 1250, loss[loss=0.2707, simple_loss=0.3344, pruned_loss=0.1035, over 7927.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.316, pruned_loss=0.08425, over 1614481.02 frames. ], batch size: 20, lr: 7.82e-03, grad_scale: 8.0 +2023-02-06 08:20:39,874 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74026.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:20:46,794 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.79 vs. limit=5.0 +2023-02-06 08:20:51,572 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 2.519e+02 3.075e+02 3.983e+02 7.817e+02, threshold=6.150e+02, percent-clipped=4.0 +2023-02-06 08:20:54,942 INFO [train.py:901] (1/4) Epoch 10, batch 1300, loss[loss=0.2775, simple_loss=0.3483, pruned_loss=0.1033, over 7149.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3155, pruned_loss=0.0843, over 1611571.88 frames. ], batch size: 72, lr: 7.81e-03, grad_scale: 8.0 +2023-02-06 08:20:57,208 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74051.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:20:58,728 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 08:21:07,834 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74067.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:21:19,111 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:21:21,530 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.85 vs. limit=5.0 +2023-02-06 08:21:26,841 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:21:29,354 INFO [train.py:901] (1/4) Epoch 10, batch 1350, loss[loss=0.2159, simple_loss=0.2904, pruned_loss=0.07071, over 7704.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3163, pruned_loss=0.08488, over 1612123.63 frames. ], batch size: 18, lr: 7.81e-03, grad_scale: 8.0 +2023-02-06 08:21:59,537 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 2.773e+02 3.448e+02 4.052e+02 8.675e+02, threshold=6.895e+02, percent-clipped=5.0 +2023-02-06 08:22:02,261 INFO [train.py:901] (1/4) Epoch 10, batch 1400, loss[loss=0.2363, simple_loss=0.3061, pruned_loss=0.08321, over 8137.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3164, pruned_loss=0.08512, over 1613194.86 frames. ], batch size: 22, lr: 7.81e-03, grad_scale: 8.0 +2023-02-06 08:22:25,376 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.48 vs. limit=5.0 +2023-02-06 08:22:38,034 INFO [train.py:901] (1/4) Epoch 10, batch 1450, loss[loss=0.2123, simple_loss=0.282, pruned_loss=0.07133, over 7799.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3154, pruned_loss=0.08398, over 1617248.31 frames. ], batch size: 19, lr: 7.81e-03, grad_scale: 8.0 +2023-02-06 08:22:41,677 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 08:22:45,869 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74209.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:23:09,188 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.838e+02 2.525e+02 3.045e+02 3.954e+02 1.310e+03, threshold=6.089e+02, percent-clipped=4.0 +2023-02-06 08:23:11,856 INFO [train.py:901] (1/4) Epoch 10, batch 1500, loss[loss=0.253, simple_loss=0.3301, pruned_loss=0.08798, over 8364.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3146, pruned_loss=0.08376, over 1613686.78 frames. ], batch size: 24, lr: 7.80e-03, grad_scale: 8.0 +2023-02-06 08:23:18,344 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74258.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:23:18,741 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-06 08:23:19,294 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.50 vs. limit=5.0 +2023-02-06 08:23:19,838 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-06 08:23:46,603 INFO [train.py:901] (1/4) Epoch 10, batch 1550, loss[loss=0.2093, simple_loss=0.2812, pruned_loss=0.06875, over 7660.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3137, pruned_loss=0.08316, over 1612928.77 frames. ], batch size: 19, lr: 7.80e-03, grad_scale: 8.0 +2023-02-06 08:24:05,702 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74323.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:24:15,645 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 08:24:19,871 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.540e+02 3.095e+02 3.981e+02 6.537e+02, threshold=6.190e+02, percent-clipped=3.0 +2023-02-06 08:24:22,714 INFO [train.py:901] (1/4) Epoch 10, batch 1600, loss[loss=0.2099, simple_loss=0.2807, pruned_loss=0.06956, over 7274.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3153, pruned_loss=0.08396, over 1616258.11 frames. ], batch size: 16, lr: 7.80e-03, grad_scale: 8.0 +2023-02-06 08:24:22,924 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74348.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:24:56,950 INFO [train.py:901] (1/4) Epoch 10, batch 1650, loss[loss=0.2609, simple_loss=0.336, pruned_loss=0.09289, over 8317.00 frames. ], tot_loss[loss=0.241, simple_loss=0.315, pruned_loss=0.08353, over 1614241.20 frames. ], batch size: 49, lr: 7.79e-03, grad_scale: 8.0 +2023-02-06 08:25:18,080 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74426.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:25:30,269 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.508e+02 3.008e+02 3.971e+02 8.483e+02, threshold=6.016e+02, percent-clipped=6.0 +2023-02-06 08:25:31,723 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7980, 2.2402, 1.6543, 2.6414, 1.1793, 1.2654, 1.7410, 2.0087], + device='cuda:1'), covar=tensor([0.0851, 0.0695, 0.1213, 0.0449, 0.1222, 0.1790, 0.1033, 0.0786], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0216, 0.0261, 0.0217, 0.0221, 0.0255, 0.0261, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 08:25:32,864 INFO [train.py:901] (1/4) Epoch 10, batch 1700, loss[loss=0.2308, simple_loss=0.3014, pruned_loss=0.08006, over 7549.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3148, pruned_loss=0.08385, over 1610834.05 frames. ], batch size: 18, lr: 7.79e-03, grad_scale: 8.0 +2023-02-06 08:25:44,197 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74465.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:26:00,777 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:26:05,899 INFO [train.py:901] (1/4) Epoch 10, batch 1750, loss[loss=0.2631, simple_loss=0.3389, pruned_loss=0.09367, over 8316.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3159, pruned_loss=0.08444, over 1613637.43 frames. ], batch size: 25, lr: 7.79e-03, grad_scale: 8.0 +2023-02-06 08:26:36,855 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74541.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:26:38,766 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.804e+02 3.517e+02 4.698e+02 1.546e+03, threshold=7.034e+02, percent-clipped=7.0 +2023-02-06 08:26:41,533 INFO [train.py:901] (1/4) Epoch 10, batch 1800, loss[loss=0.2325, simple_loss=0.303, pruned_loss=0.08099, over 7963.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3159, pruned_loss=0.08455, over 1611798.24 frames. ], batch size: 21, lr: 7.79e-03, grad_scale: 8.0 +2023-02-06 08:27:14,907 INFO [train.py:901] (1/4) Epoch 10, batch 1850, loss[loss=0.233, simple_loss=0.3065, pruned_loss=0.07971, over 8362.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3151, pruned_loss=0.08451, over 1610602.40 frames. ], batch size: 24, lr: 7.78e-03, grad_scale: 8.0 +2023-02-06 08:27:17,786 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74602.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:27:25,276 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74613.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:27:46,972 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 2.722e+02 3.219e+02 4.226e+02 1.097e+03, threshold=6.437e+02, percent-clipped=2.0 +2023-02-06 08:27:50,407 INFO [train.py:901] (1/4) Epoch 10, batch 1900, loss[loss=0.2705, simple_loss=0.3373, pruned_loss=0.1018, over 8669.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3161, pruned_loss=0.08487, over 1611227.67 frames. ], batch size: 39, lr: 7.78e-03, grad_scale: 8.0 +2023-02-06 08:28:13,881 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 08:28:25,266 INFO [train.py:901] (1/4) Epoch 10, batch 1950, loss[loss=0.2746, simple_loss=0.3406, pruned_loss=0.1043, over 6794.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3167, pruned_loss=0.08533, over 1613937.10 frames. ], batch size: 71, lr: 7.78e-03, grad_scale: 8.0 +2023-02-06 08:28:25,962 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 08:28:34,313 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.58 vs. limit=5.0 +2023-02-06 08:28:38,209 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74717.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:28:43,960 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 08:28:56,735 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.465e+02 3.030e+02 3.717e+02 6.494e+02, threshold=6.060e+02, percent-clipped=3.0 +2023-02-06 08:28:59,502 INFO [train.py:901] (1/4) Epoch 10, batch 2000, loss[loss=0.2786, simple_loss=0.3431, pruned_loss=0.107, over 8557.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.316, pruned_loss=0.08447, over 1615962.03 frames. ], batch size: 49, lr: 7.78e-03, grad_scale: 8.0 +2023-02-06 08:29:04,983 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9182, 3.9337, 2.3729, 2.5318, 2.6047, 1.8491, 2.7732, 2.9731], + device='cuda:1'), covar=tensor([0.1497, 0.0237, 0.0920, 0.0796, 0.0712, 0.1255, 0.0943, 0.0994], + device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0233, 0.0312, 0.0298, 0.0305, 0.0318, 0.0337, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 08:29:34,151 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74797.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:29:34,591 INFO [train.py:901] (1/4) Epoch 10, batch 2050, loss[loss=0.2342, simple_loss=0.3101, pruned_loss=0.07914, over 8502.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3147, pruned_loss=0.08367, over 1614410.87 frames. ], batch size: 28, lr: 7.77e-03, grad_scale: 8.0 +2023-02-06 08:29:40,112 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5128, 1.8783, 1.8712, 1.2401, 1.9995, 1.3491, 0.4707, 1.5847], + device='cuda:1'), covar=tensor([0.0302, 0.0166, 0.0150, 0.0270, 0.0194, 0.0518, 0.0481, 0.0153], + device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0309, 0.0266, 0.0367, 0.0296, 0.0456, 0.0349, 0.0334], + device='cuda:1'), out_proj_covar=tensor([1.1003e-04, 8.8041e-05, 7.6442e-05, 1.0546e-04, 8.6185e-05, 1.4293e-04, + 1.0245e-04, 9.7077e-05], device='cuda:1') +2023-02-06 08:29:50,349 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74822.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:30:04,620 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.471e+02 3.084e+02 4.282e+02 1.276e+03, threshold=6.169e+02, percent-clipped=5.0 +2023-02-06 08:30:07,347 INFO [train.py:901] (1/4) Epoch 10, batch 2100, loss[loss=0.2096, simple_loss=0.2977, pruned_loss=0.06078, over 8456.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3131, pruned_loss=0.08303, over 1616130.77 frames. ], batch size: 25, lr: 7.77e-03, grad_scale: 8.0 +2023-02-06 08:30:42,984 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 08:30:43,219 INFO [train.py:901] (1/4) Epoch 10, batch 2150, loss[loss=0.2221, simple_loss=0.2897, pruned_loss=0.07724, over 7532.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3127, pruned_loss=0.08281, over 1611211.61 frames. ], batch size: 18, lr: 7.77e-03, grad_scale: 8.0 +2023-02-06 08:31:02,634 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:31:13,920 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.764e+02 2.626e+02 3.226e+02 3.775e+02 6.882e+02, threshold=6.451e+02, percent-clipped=1.0 +2023-02-06 08:31:16,699 INFO [train.py:901] (1/4) Epoch 10, batch 2200, loss[loss=0.2584, simple_loss=0.3209, pruned_loss=0.0979, over 7694.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3135, pruned_loss=0.08323, over 1613163.28 frames. ], batch size: 18, lr: 7.77e-03, grad_scale: 8.0 +2023-02-06 08:31:22,705 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74957.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:31:29,470 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4834, 2.0029, 3.1004, 2.4431, 2.7131, 2.1941, 1.7225, 1.4538], + device='cuda:1'), covar=tensor([0.3408, 0.3655, 0.1031, 0.2252, 0.1753, 0.2001, 0.1636, 0.3854], + device='cuda:1'), in_proj_covar=tensor([0.0868, 0.0841, 0.0718, 0.0825, 0.0917, 0.0775, 0.0696, 0.0756], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:31:33,572 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74973.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:31:39,088 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.57 vs. limit=5.0 +2023-02-06 08:31:39,394 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74982.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:31:50,364 INFO [train.py:901] (1/4) Epoch 10, batch 2250, loss[loss=0.3015, simple_loss=0.366, pruned_loss=0.1186, over 8251.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3133, pruned_loss=0.083, over 1612169.37 frames. ], batch size: 24, lr: 7.76e-03, grad_scale: 8.0 +2023-02-06 08:31:50,514 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1486, 1.1363, 3.3211, 1.0213, 2.8269, 2.8114, 2.9969, 2.8893], + device='cuda:1'), covar=tensor([0.0685, 0.3885, 0.0727, 0.3316, 0.1434, 0.0949, 0.0665, 0.0808], + device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0560, 0.0560, 0.0514, 0.0587, 0.0506, 0.0493, 0.0557], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:31:50,579 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74998.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:32:04,484 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1846, 1.4523, 1.5642, 1.3391, 1.1108, 1.3306, 1.7952, 1.6858], + device='cuda:1'), covar=tensor([0.0512, 0.1353, 0.1802, 0.1478, 0.0573, 0.1615, 0.0704, 0.0608], + device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0156, 0.0196, 0.0161, 0.0106, 0.0166, 0.0118, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 08:32:23,118 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.748e+02 3.468e+02 4.709e+02 1.048e+03, threshold=6.936e+02, percent-clipped=3.0 +2023-02-06 08:32:25,869 INFO [train.py:901] (1/4) Epoch 10, batch 2300, loss[loss=0.2096, simple_loss=0.2872, pruned_loss=0.06601, over 7258.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3134, pruned_loss=0.08304, over 1611988.30 frames. ], batch size: 16, lr: 7.76e-03, grad_scale: 8.0 +2023-02-06 08:32:42,307 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75072.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:32:59,565 INFO [train.py:901] (1/4) Epoch 10, batch 2350, loss[loss=0.2365, simple_loss=0.3237, pruned_loss=0.07468, over 8327.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.3141, pruned_loss=0.08333, over 1610947.95 frames. ], batch size: 25, lr: 7.76e-03, grad_scale: 8.0 +2023-02-06 08:33:01,993 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 08:33:33,026 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.526e+02 3.215e+02 4.182e+02 1.054e+03, threshold=6.430e+02, percent-clipped=5.0 +2023-02-06 08:33:35,803 INFO [train.py:901] (1/4) Epoch 10, batch 2400, loss[loss=0.22, simple_loss=0.2835, pruned_loss=0.07829, over 7783.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3153, pruned_loss=0.08406, over 1614306.27 frames. ], batch size: 19, lr: 7.76e-03, grad_scale: 8.0 +2023-02-06 08:34:08,760 INFO [train.py:901] (1/4) Epoch 10, batch 2450, loss[loss=0.1981, simple_loss=0.2749, pruned_loss=0.06066, over 7528.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.315, pruned_loss=0.08433, over 1610719.33 frames. ], batch size: 18, lr: 7.75e-03, grad_scale: 8.0 +2023-02-06 08:34:40,821 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.699e+02 2.613e+02 3.092e+02 4.227e+02 1.037e+03, threshold=6.184e+02, percent-clipped=5.0 +2023-02-06 08:34:43,535 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75246.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:34:44,711 INFO [train.py:901] (1/4) Epoch 10, batch 2500, loss[loss=0.1932, simple_loss=0.2705, pruned_loss=0.05796, over 7792.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3144, pruned_loss=0.08376, over 1610098.57 frames. ], batch size: 19, lr: 7.75e-03, grad_scale: 8.0 +2023-02-06 08:35:00,231 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75271.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:35:11,667 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:35:18,083 INFO [train.py:901] (1/4) Epoch 10, batch 2550, loss[loss=0.239, simple_loss=0.3202, pruned_loss=0.07891, over 8029.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3152, pruned_loss=0.08418, over 1613711.06 frames. ], batch size: 22, lr: 7.75e-03, grad_scale: 8.0 +2023-02-06 08:35:33,987 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0633, 1.1940, 4.2999, 1.5906, 3.7166, 3.4886, 3.8271, 3.6874], + device='cuda:1'), covar=tensor([0.0495, 0.4200, 0.0452, 0.3238, 0.1141, 0.0802, 0.0553, 0.0630], + device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0559, 0.0566, 0.0516, 0.0589, 0.0506, 0.0492, 0.0555], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:35:36,578 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75326.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:35:38,119 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:35:49,133 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.664e+02 3.245e+02 3.791e+02 6.757e+02, threshold=6.490e+02, percent-clipped=2.0 +2023-02-06 08:35:51,833 INFO [train.py:901] (1/4) Epoch 10, batch 2600, loss[loss=0.2282, simple_loss=0.3178, pruned_loss=0.06934, over 8456.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3155, pruned_loss=0.08469, over 1609253.91 frames. ], batch size: 25, lr: 7.75e-03, grad_scale: 8.0 +2023-02-06 08:35:55,398 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75353.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:36:01,374 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7789, 4.2057, 2.6445, 2.7941, 2.7774, 1.8526, 2.8662, 3.2035], + device='cuda:1'), covar=tensor([0.1550, 0.0244, 0.0892, 0.0710, 0.0806, 0.1532, 0.1064, 0.0834], + device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0236, 0.0314, 0.0299, 0.0308, 0.0322, 0.0341, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 08:36:08,167 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8295, 2.4571, 4.7858, 1.3809, 3.1194, 2.3676, 1.8918, 2.9872], + device='cuda:1'), covar=tensor([0.1506, 0.2082, 0.0514, 0.3558, 0.1432, 0.2530, 0.1525, 0.1975], + device='cuda:1'), in_proj_covar=tensor([0.0479, 0.0498, 0.0532, 0.0566, 0.0605, 0.0544, 0.0464, 0.0605], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 08:36:19,401 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75386.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:36:27,353 INFO [train.py:901] (1/4) Epoch 10, batch 2650, loss[loss=0.2887, simple_loss=0.3586, pruned_loss=0.1094, over 8625.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3163, pruned_loss=0.08536, over 1607998.16 frames. ], batch size: 31, lr: 7.74e-03, grad_scale: 8.0 +2023-02-06 08:36:56,762 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75441.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:36:58,553 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 2.723e+02 3.413e+02 4.384e+02 8.455e+02, threshold=6.827e+02, percent-clipped=3.0 +2023-02-06 08:37:01,358 INFO [train.py:901] (1/4) Epoch 10, batch 2700, loss[loss=0.2553, simple_loss=0.3336, pruned_loss=0.08849, over 8716.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3155, pruned_loss=0.08462, over 1604851.44 frames. ], batch size: 30, lr: 7.74e-03, grad_scale: 8.0 +2023-02-06 08:37:37,687 INFO [train.py:901] (1/4) Epoch 10, batch 2750, loss[loss=0.2839, simple_loss=0.3487, pruned_loss=0.1096, over 8502.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3173, pruned_loss=0.08515, over 1611217.63 frames. ], batch size: 26, lr: 7.74e-03, grad_scale: 8.0 +2023-02-06 08:37:47,101 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5272, 1.5584, 1.7280, 1.4580, 1.0896, 1.7884, 0.0693, 1.1376], + device='cuda:1'), covar=tensor([0.3100, 0.1640, 0.0564, 0.1554, 0.4377, 0.0555, 0.3474, 0.2235], + device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0165, 0.0094, 0.0217, 0.0258, 0.0099, 0.0164, 0.0161], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 08:37:55,834 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2937, 1.9638, 2.9877, 2.3352, 2.6534, 2.0765, 1.6397, 1.4666], + device='cuda:1'), covar=tensor([0.3454, 0.3678, 0.1003, 0.2220, 0.1716, 0.2119, 0.1692, 0.3993], + device='cuda:1'), in_proj_covar=tensor([0.0863, 0.0840, 0.0711, 0.0823, 0.0920, 0.0779, 0.0697, 0.0755], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:38:01,676 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7806, 2.0334, 1.7481, 2.6600, 1.2581, 1.3849, 1.7683, 1.9889], + device='cuda:1'), covar=tensor([0.0931, 0.0984, 0.1219, 0.0408, 0.1349, 0.1688, 0.1061, 0.0875], + device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0220, 0.0265, 0.0220, 0.0225, 0.0257, 0.0265, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 08:38:08,174 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 2.609e+02 3.111e+02 3.957e+02 1.084e+03, threshold=6.223e+02, percent-clipped=3.0 +2023-02-06 08:38:10,722 INFO [train.py:901] (1/4) Epoch 10, batch 2800, loss[loss=0.2763, simple_loss=0.3494, pruned_loss=0.1016, over 8498.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3174, pruned_loss=0.08556, over 1612237.46 frames. ], batch size: 26, lr: 7.74e-03, grad_scale: 8.0 +2023-02-06 08:38:13,273 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 08:38:39,098 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75590.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:38:44,121 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.68 vs. limit=5.0 +2023-02-06 08:38:45,064 INFO [train.py:901] (1/4) Epoch 10, batch 2850, loss[loss=0.2338, simple_loss=0.3107, pruned_loss=0.07842, over 8127.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.317, pruned_loss=0.0849, over 1612672.17 frames. ], batch size: 22, lr: 7.73e-03, grad_scale: 8.0 +2023-02-06 08:39:04,152 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.76 vs. limit=5.0 +2023-02-06 08:39:09,363 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75632.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:39:16,267 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75642.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:39:17,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 2.670e+02 3.172e+02 3.749e+02 6.038e+02, threshold=6.343e+02, percent-clipped=0.0 +2023-02-06 08:39:20,068 INFO [train.py:901] (1/4) Epoch 10, batch 2900, loss[loss=0.2479, simple_loss=0.3331, pruned_loss=0.08128, over 8358.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3176, pruned_loss=0.08557, over 1614469.66 frames. ], batch size: 24, lr: 7.73e-03, grad_scale: 8.0 +2023-02-06 08:39:32,781 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:39:49,788 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 08:39:53,283 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75697.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:39:53,745 INFO [train.py:901] (1/4) Epoch 10, batch 2950, loss[loss=0.2427, simple_loss=0.3271, pruned_loss=0.07912, over 8203.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3167, pruned_loss=0.08506, over 1615443.09 frames. ], batch size: 23, lr: 7.73e-03, grad_scale: 16.0 +2023-02-06 08:39:58,777 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75705.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:40:11,379 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75722.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:40:26,685 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.480e+02 3.030e+02 3.596e+02 1.304e+03, threshold=6.060e+02, percent-clipped=4.0 +2023-02-06 08:40:28,962 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75747.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:40:29,496 INFO [train.py:901] (1/4) Epoch 10, batch 3000, loss[loss=0.2136, simple_loss=0.2868, pruned_loss=0.0702, over 8144.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3152, pruned_loss=0.08421, over 1614012.68 frames. ], batch size: 22, lr: 7.73e-03, grad_scale: 16.0 +2023-02-06 08:40:29,497 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 08:40:41,877 INFO [train.py:935] (1/4) Epoch 10, validation: loss=0.1918, simple_loss=0.2916, pruned_loss=0.04599, over 944034.00 frames. +2023-02-06 08:40:41,878 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 08:41:15,547 INFO [train.py:901] (1/4) Epoch 10, batch 3050, loss[loss=0.2179, simple_loss=0.2952, pruned_loss=0.07031, over 6332.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3154, pruned_loss=0.08445, over 1608253.47 frames. ], batch size: 14, lr: 7.72e-03, grad_scale: 16.0 +2023-02-06 08:41:47,924 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.518e+02 3.138e+02 4.468e+02 1.006e+03, threshold=6.276e+02, percent-clipped=13.0 +2023-02-06 08:41:50,032 INFO [train.py:901] (1/4) Epoch 10, batch 3100, loss[loss=0.1979, simple_loss=0.2821, pruned_loss=0.05683, over 8042.00 frames. ], tot_loss[loss=0.242, simple_loss=0.315, pruned_loss=0.08448, over 1603066.99 frames. ], batch size: 22, lr: 7.72e-03, grad_scale: 8.0 +2023-02-06 08:41:52,187 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2804, 1.5852, 4.4959, 1.6824, 3.8291, 3.7284, 4.0096, 3.8649], + device='cuda:1'), covar=tensor([0.0583, 0.3923, 0.0497, 0.3207, 0.1287, 0.0797, 0.0556, 0.0645], + device='cuda:1'), in_proj_covar=tensor([0.0458, 0.0564, 0.0578, 0.0525, 0.0602, 0.0512, 0.0499, 0.0565], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:42:25,561 INFO [train.py:901] (1/4) Epoch 10, batch 3150, loss[loss=0.2446, simple_loss=0.3166, pruned_loss=0.08632, over 8034.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3169, pruned_loss=0.08568, over 1608365.25 frames. ], batch size: 22, lr: 7.72e-03, grad_scale: 8.0 +2023-02-06 08:42:48,061 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6378, 1.3286, 4.8119, 1.7787, 4.1992, 3.9804, 4.3359, 4.2214], + device='cuda:1'), covar=tensor([0.0459, 0.4198, 0.0389, 0.3128, 0.0959, 0.0745, 0.0462, 0.0506], + device='cuda:1'), in_proj_covar=tensor([0.0455, 0.0562, 0.0572, 0.0521, 0.0599, 0.0510, 0.0497, 0.0560], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:42:57,481 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 2.584e+02 3.323e+02 3.941e+02 8.938e+02, threshold=6.646e+02, percent-clipped=3.0 +2023-02-06 08:42:59,544 INFO [train.py:901] (1/4) Epoch 10, batch 3200, loss[loss=0.2428, simple_loss=0.3238, pruned_loss=0.08089, over 8475.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3163, pruned_loss=0.08533, over 1604589.59 frames. ], batch size: 29, lr: 7.72e-03, grad_scale: 8.0 +2023-02-06 08:43:09,326 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75961.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:43:28,331 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75986.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:43:28,980 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8563, 3.8554, 2.3852, 2.6520, 2.8999, 1.8977, 2.8386, 2.9571], + device='cuda:1'), covar=tensor([0.1695, 0.0333, 0.0982, 0.0782, 0.0716, 0.1342, 0.1032, 0.1136], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0237, 0.0314, 0.0301, 0.0312, 0.0326, 0.0343, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 08:43:36,371 INFO [train.py:901] (1/4) Epoch 10, batch 3250, loss[loss=0.2256, simple_loss=0.3143, pruned_loss=0.06846, over 8110.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3159, pruned_loss=0.08462, over 1608241.17 frames. ], batch size: 23, lr: 7.71e-03, grad_scale: 8.0 +2023-02-06 08:43:41,259 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76003.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:43:57,878 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76028.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:44:04,546 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7292, 1.5511, 2.6400, 1.1691, 2.0014, 2.7998, 3.0712, 1.9929], + device='cuda:1'), covar=tensor([0.1252, 0.1394, 0.0585, 0.2403, 0.0951, 0.0500, 0.0646, 0.1290], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0294, 0.0254, 0.0284, 0.0270, 0.0234, 0.0319, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 08:44:09,017 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.688e+02 3.300e+02 3.989e+02 9.835e+02, threshold=6.601e+02, percent-clipped=4.0 +2023-02-06 08:44:11,020 INFO [train.py:901] (1/4) Epoch 10, batch 3300, loss[loss=0.2341, simple_loss=0.3076, pruned_loss=0.08037, over 7706.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3154, pruned_loss=0.08442, over 1607341.34 frames. ], batch size: 18, lr: 7.71e-03, grad_scale: 8.0 +2023-02-06 08:44:11,218 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3564, 2.6255, 1.8660, 2.1717, 2.2484, 1.4595, 2.0910, 2.2126], + device='cuda:1'), covar=tensor([0.1357, 0.0330, 0.0993, 0.0517, 0.0626, 0.1474, 0.0853, 0.0814], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0235, 0.0313, 0.0298, 0.0310, 0.0324, 0.0338, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 08:44:47,579 INFO [train.py:901] (1/4) Epoch 10, batch 3350, loss[loss=0.2476, simple_loss=0.3055, pruned_loss=0.0948, over 7527.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3154, pruned_loss=0.08427, over 1608681.01 frames. ], batch size: 18, lr: 7.71e-03, grad_scale: 8.0 +2023-02-06 08:44:58,400 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4802, 1.5052, 4.2515, 1.9525, 2.3222, 4.8177, 4.7341, 4.0719], + device='cuda:1'), covar=tensor([0.0949, 0.1600, 0.0312, 0.1871, 0.1134, 0.0180, 0.0348, 0.0588], + device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0293, 0.0252, 0.0283, 0.0268, 0.0233, 0.0318, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 08:45:18,634 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.652e+02 3.239e+02 4.192e+02 7.352e+02, threshold=6.477e+02, percent-clipped=1.0 +2023-02-06 08:45:20,670 INFO [train.py:901] (1/4) Epoch 10, batch 3400, loss[loss=0.2662, simple_loss=0.33, pruned_loss=0.1012, over 8193.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3155, pruned_loss=0.08451, over 1608553.51 frames. ], batch size: 23, lr: 7.71e-03, grad_scale: 8.0 +2023-02-06 08:45:55,814 INFO [train.py:901] (1/4) Epoch 10, batch 3450, loss[loss=0.2793, simple_loss=0.3381, pruned_loss=0.1103, over 8566.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3156, pruned_loss=0.08478, over 1607338.84 frames. ], batch size: 31, lr: 7.70e-03, grad_scale: 8.0 +2023-02-06 08:46:30,138 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.394e+02 3.045e+02 3.881e+02 9.338e+02, threshold=6.090e+02, percent-clipped=3.0 +2023-02-06 08:46:32,227 INFO [train.py:901] (1/4) Epoch 10, batch 3500, loss[loss=0.2622, simple_loss=0.3326, pruned_loss=0.09589, over 8362.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3156, pruned_loss=0.0844, over 1612994.98 frames. ], batch size: 24, lr: 7.70e-03, grad_scale: 8.0 +2023-02-06 08:46:48,064 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 08:46:59,315 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76287.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:47:06,734 INFO [train.py:901] (1/4) Epoch 10, batch 3550, loss[loss=0.2377, simple_loss=0.2868, pruned_loss=0.09425, over 7712.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3158, pruned_loss=0.08485, over 1608426.94 frames. ], batch size: 18, lr: 7.70e-03, grad_scale: 8.0 +2023-02-06 08:47:17,273 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:47:39,382 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76341.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:47:42,024 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 2.725e+02 3.470e+02 4.316e+02 7.747e+02, threshold=6.941e+02, percent-clipped=6.0 +2023-02-06 08:47:44,159 INFO [train.py:901] (1/4) Epoch 10, batch 3600, loss[loss=0.2557, simple_loss=0.33, pruned_loss=0.0907, over 8457.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.3148, pruned_loss=0.08404, over 1613342.46 frames. ], batch size: 27, lr: 7.70e-03, grad_scale: 8.0 +2023-02-06 08:47:44,401 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5253, 1.9601, 2.0319, 1.2353, 2.1470, 1.3274, 0.7276, 1.7079], + device='cuda:1'), covar=tensor([0.0493, 0.0243, 0.0216, 0.0404, 0.0291, 0.0671, 0.0588, 0.0221], + device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0308, 0.0264, 0.0370, 0.0298, 0.0458, 0.0353, 0.0336], + device='cuda:1'), out_proj_covar=tensor([1.0895e-04, 8.7400e-05, 7.5535e-05, 1.0599e-04, 8.6524e-05, 1.4287e-04, + 1.0330e-04, 9.7498e-05], device='cuda:1') +2023-02-06 08:48:18,378 INFO [train.py:901] (1/4) Epoch 10, batch 3650, loss[loss=0.1968, simple_loss=0.2821, pruned_loss=0.05573, over 7976.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3146, pruned_loss=0.08329, over 1618082.52 frames. ], batch size: 21, lr: 7.69e-03, grad_scale: 8.0 +2023-02-06 08:48:50,989 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.530e+02 3.057e+02 3.624e+02 8.995e+02, threshold=6.114e+02, percent-clipped=3.0 +2023-02-06 08:48:52,986 INFO [train.py:901] (1/4) Epoch 10, batch 3700, loss[loss=0.2171, simple_loss=0.3041, pruned_loss=0.06508, over 7803.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3142, pruned_loss=0.08312, over 1620679.70 frames. ], batch size: 20, lr: 7.69e-03, grad_scale: 8.0 +2023-02-06 08:48:54,935 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 08:49:11,482 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2257, 1.7813, 2.8854, 2.2901, 2.5403, 2.0740, 1.6294, 1.1244], + device='cuda:1'), covar=tensor([0.3791, 0.4027, 0.0949, 0.2380, 0.1824, 0.2241, 0.1796, 0.4149], + device='cuda:1'), in_proj_covar=tensor([0.0873, 0.0846, 0.0711, 0.0829, 0.0920, 0.0779, 0.0699, 0.0757], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:49:28,892 INFO [train.py:901] (1/4) Epoch 10, batch 3750, loss[loss=0.2397, simple_loss=0.316, pruned_loss=0.08171, over 8079.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3151, pruned_loss=0.08299, over 1623679.37 frames. ], batch size: 21, lr: 7.69e-03, grad_scale: 8.0 +2023-02-06 08:50:00,300 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.765e+02 3.416e+02 4.278e+02 1.031e+03, threshold=6.832e+02, percent-clipped=4.0 +2023-02-06 08:50:02,988 INFO [train.py:901] (1/4) Epoch 10, batch 3800, loss[loss=0.2859, simple_loss=0.3544, pruned_loss=0.1087, over 8126.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3146, pruned_loss=0.08286, over 1618901.52 frames. ], batch size: 22, lr: 7.69e-03, grad_scale: 8.0 +2023-02-06 08:50:38,444 INFO [train.py:901] (1/4) Epoch 10, batch 3850, loss[loss=0.2206, simple_loss=0.3103, pruned_loss=0.06548, over 8261.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3146, pruned_loss=0.08321, over 1617718.85 frames. ], batch size: 24, lr: 7.68e-03, grad_scale: 8.0 +2023-02-06 08:50:53,209 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9582, 1.6769, 1.3550, 1.6158, 1.2541, 1.1256, 1.2371, 1.3783], + device='cuda:1'), covar=tensor([0.0989, 0.0434, 0.1068, 0.0473, 0.0719, 0.1381, 0.0845, 0.0686], + device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0234, 0.0309, 0.0295, 0.0303, 0.0322, 0.0337, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 08:50:59,040 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 08:51:00,491 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76631.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:51:09,918 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.563e+02 3.093e+02 4.191e+02 1.151e+03, threshold=6.187e+02, percent-clipped=5.0 +2023-02-06 08:51:11,982 INFO [train.py:901] (1/4) Epoch 10, batch 3900, loss[loss=0.2063, simple_loss=0.276, pruned_loss=0.06827, over 7645.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3134, pruned_loss=0.08268, over 1618257.44 frames. ], batch size: 19, lr: 7.68e-03, grad_scale: 8.0 +2023-02-06 08:51:17,435 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76656.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:51:38,044 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76685.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:51:48,063 INFO [train.py:901] (1/4) Epoch 10, batch 3950, loss[loss=0.2446, simple_loss=0.3103, pruned_loss=0.08948, over 7811.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3145, pruned_loss=0.08399, over 1614728.80 frames. ], batch size: 20, lr: 7.68e-03, grad_scale: 8.0 +2023-02-06 08:52:19,558 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.430e+02 3.097e+02 3.693e+02 7.444e+02, threshold=6.193e+02, percent-clipped=4.0 +2023-02-06 08:52:20,445 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76746.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:52:21,615 INFO [train.py:901] (1/4) Epoch 10, batch 4000, loss[loss=0.2382, simple_loss=0.3174, pruned_loss=0.07955, over 8623.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3143, pruned_loss=0.08396, over 1610347.33 frames. ], batch size: 34, lr: 7.68e-03, grad_scale: 8.0 +2023-02-06 08:52:32,221 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 08:52:37,417 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76771.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:52:51,544 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1689, 1.7973, 2.7063, 2.1317, 2.3482, 2.0593, 1.6669, 0.9460], + device='cuda:1'), covar=tensor([0.3470, 0.3510, 0.0941, 0.2148, 0.1714, 0.2067, 0.1599, 0.3756], + device='cuda:1'), in_proj_covar=tensor([0.0865, 0.0838, 0.0707, 0.0823, 0.0912, 0.0773, 0.0693, 0.0750], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:52:56,078 INFO [train.py:901] (1/4) Epoch 10, batch 4050, loss[loss=0.2407, simple_loss=0.315, pruned_loss=0.08318, over 8443.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3155, pruned_loss=0.08479, over 1609634.85 frames. ], batch size: 49, lr: 7.67e-03, grad_scale: 8.0 +2023-02-06 08:52:57,682 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76800.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:53:27,753 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.17 vs. limit=2.0 +2023-02-06 08:53:29,318 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.637e+02 3.294e+02 4.061e+02 9.505e+02, threshold=6.587e+02, percent-clipped=7.0 +2023-02-06 08:53:31,229 INFO [train.py:901] (1/4) Epoch 10, batch 4100, loss[loss=0.2385, simple_loss=0.3192, pruned_loss=0.07889, over 8332.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3167, pruned_loss=0.0855, over 1611277.40 frames. ], batch size: 25, lr: 7.67e-03, grad_scale: 8.0 +2023-02-06 08:53:31,386 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6611, 1.4256, 4.8352, 1.7381, 4.2888, 4.0772, 4.4289, 4.1951], + device='cuda:1'), covar=tensor([0.0422, 0.3913, 0.0333, 0.2922, 0.0842, 0.0718, 0.0400, 0.0554], + device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0552, 0.0556, 0.0513, 0.0582, 0.0499, 0.0489, 0.0552], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:53:37,247 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76857.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:54:04,759 INFO [train.py:901] (1/4) Epoch 10, batch 4150, loss[loss=0.2674, simple_loss=0.3449, pruned_loss=0.09493, over 8498.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3147, pruned_loss=0.08429, over 1608914.32 frames. ], batch size: 26, lr: 7.67e-03, grad_scale: 8.0 +2023-02-06 08:54:15,625 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1921, 1.1344, 3.3377, 0.9878, 2.8849, 2.8329, 3.0365, 2.9044], + device='cuda:1'), covar=tensor([0.0699, 0.3625, 0.0702, 0.3132, 0.1396, 0.1018, 0.0668, 0.0904], + device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0555, 0.0559, 0.0515, 0.0585, 0.0501, 0.0490, 0.0557], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 08:54:38,768 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.554e+02 2.505e+02 2.967e+02 3.617e+02 8.554e+02, threshold=5.933e+02, percent-clipped=2.0 +2023-02-06 08:54:39,589 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7944, 3.6751, 3.4206, 1.8236, 3.2978, 3.3231, 3.4773, 3.0633], + device='cuda:1'), covar=tensor([0.0987, 0.0828, 0.1174, 0.4938, 0.1018, 0.1078, 0.1446, 0.1090], + device='cuda:1'), in_proj_covar=tensor([0.0451, 0.0362, 0.0372, 0.0471, 0.0366, 0.0357, 0.0364, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 08:54:40,856 INFO [train.py:901] (1/4) Epoch 10, batch 4200, loss[loss=0.2038, simple_loss=0.2779, pruned_loss=0.06482, over 7423.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3136, pruned_loss=0.08407, over 1602992.93 frames. ], batch size: 17, lr: 7.67e-03, grad_scale: 8.0 +2023-02-06 08:54:47,338 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.56 vs. limit=5.0 +2023-02-06 08:55:00,918 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 08:55:14,236 INFO [train.py:901] (1/4) Epoch 10, batch 4250, loss[loss=0.2596, simple_loss=0.3425, pruned_loss=0.0884, over 8527.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3135, pruned_loss=0.08353, over 1606676.16 frames. ], batch size: 28, lr: 7.66e-03, grad_scale: 8.0 +2023-02-06 08:55:17,204 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77002.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:55:23,834 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 08:55:29,245 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9896, 6.1187, 5.3553, 2.8341, 5.4267, 5.7561, 5.6253, 5.2740], + device='cuda:1'), covar=tensor([0.0552, 0.0440, 0.0847, 0.3974, 0.0644, 0.0653, 0.1165, 0.0592], + device='cuda:1'), in_proj_covar=tensor([0.0456, 0.0364, 0.0373, 0.0473, 0.0370, 0.0360, 0.0367, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 08:55:34,061 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:55:34,071 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77027.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:55:46,497 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.530e+02 3.131e+02 3.743e+02 6.568e+02, threshold=6.262e+02, percent-clipped=1.0 +2023-02-06 08:55:48,443 INFO [train.py:901] (1/4) Epoch 10, batch 4300, loss[loss=0.271, simple_loss=0.3359, pruned_loss=0.103, over 8286.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3146, pruned_loss=0.08397, over 1610538.89 frames. ], batch size: 23, lr: 7.66e-03, grad_scale: 8.0 +2023-02-06 08:55:52,732 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77052.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:55:55,460 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77056.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:55:55,973 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2082, 4.1824, 3.7687, 1.8736, 3.7507, 3.6256, 3.8825, 3.3460], + device='cuda:1'), covar=tensor([0.0811, 0.0625, 0.1011, 0.4967, 0.0890, 0.0907, 0.1185, 0.0997], + device='cuda:1'), in_proj_covar=tensor([0.0455, 0.0362, 0.0371, 0.0473, 0.0368, 0.0359, 0.0367, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 08:56:12,954 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77081.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:56:23,977 INFO [train.py:901] (1/4) Epoch 10, batch 4350, loss[loss=0.2207, simple_loss=0.2879, pruned_loss=0.07676, over 7789.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.314, pruned_loss=0.08338, over 1610134.97 frames. ], batch size: 19, lr: 7.66e-03, grad_scale: 8.0 +2023-02-06 08:56:33,992 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77113.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:56:53,833 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 08:56:54,997 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 2.751e+02 3.331e+02 4.491e+02 1.022e+03, threshold=6.663e+02, percent-clipped=8.0 +2023-02-06 08:56:57,038 INFO [train.py:901] (1/4) Epoch 10, batch 4400, loss[loss=0.2391, simple_loss=0.3227, pruned_loss=0.07777, over 8112.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3134, pruned_loss=0.08253, over 1606923.40 frames. ], batch size: 23, lr: 7.66e-03, grad_scale: 8.0 +2023-02-06 08:57:27,970 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5555, 1.8955, 2.0788, 1.0904, 2.2034, 1.5201, 0.5259, 1.7637], + device='cuda:1'), covar=tensor([0.0383, 0.0211, 0.0173, 0.0347, 0.0215, 0.0598, 0.0558, 0.0168], + device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0303, 0.0261, 0.0367, 0.0298, 0.0452, 0.0345, 0.0335], + device='cuda:1'), out_proj_covar=tensor([1.0685e-04, 8.5883e-05, 7.4300e-05, 1.0485e-04, 8.6320e-05, 1.4065e-04, + 1.0040e-04, 9.7029e-05], device='cuda:1') +2023-02-06 08:57:33,157 INFO [train.py:901] (1/4) Epoch 10, batch 4450, loss[loss=0.1629, simple_loss=0.2437, pruned_loss=0.041, over 7201.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3127, pruned_loss=0.08248, over 1603850.34 frames. ], batch size: 16, lr: 7.65e-03, grad_scale: 8.0 +2023-02-06 08:57:35,380 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77201.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:57:36,010 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 08:57:54,213 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77229.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:58:04,752 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.784e+02 3.289e+02 4.035e+02 8.452e+02, threshold=6.579e+02, percent-clipped=2.0 +2023-02-06 08:58:06,786 INFO [train.py:901] (1/4) Epoch 10, batch 4500, loss[loss=0.2994, simple_loss=0.3526, pruned_loss=0.1231, over 8364.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3125, pruned_loss=0.08237, over 1606766.39 frames. ], batch size: 24, lr: 7.65e-03, grad_scale: 8.0 +2023-02-06 08:58:27,559 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 08:58:37,991 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:58:39,438 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2001, 1.5812, 3.1995, 1.4228, 2.2354, 3.5558, 3.6019, 3.0115], + device='cuda:1'), covar=tensor([0.0797, 0.1273, 0.0339, 0.1785, 0.0865, 0.0221, 0.0428, 0.0582], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0291, 0.0251, 0.0279, 0.0265, 0.0232, 0.0321, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 08:58:41,442 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77295.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:58:43,370 INFO [train.py:901] (1/4) Epoch 10, batch 4550, loss[loss=0.1984, simple_loss=0.2793, pruned_loss=0.0587, over 7939.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3119, pruned_loss=0.08258, over 1601581.03 frames. ], batch size: 20, lr: 7.65e-03, grad_scale: 8.0 +2023-02-06 08:58:55,667 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77316.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:59:14,841 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 2.639e+02 3.213e+02 4.072e+02 8.769e+02, threshold=6.426e+02, percent-clipped=3.0 +2023-02-06 08:59:16,947 INFO [train.py:901] (1/4) Epoch 10, batch 4600, loss[loss=0.2305, simple_loss=0.3058, pruned_loss=0.07757, over 8293.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3126, pruned_loss=0.08309, over 1608410.38 frames. ], batch size: 23, lr: 7.65e-03, grad_scale: 8.0 +2023-02-06 08:59:42,598 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-06 08:59:50,937 INFO [train.py:901] (1/4) Epoch 10, batch 4650, loss[loss=0.2085, simple_loss=0.2974, pruned_loss=0.05974, over 7811.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.314, pruned_loss=0.08337, over 1610954.19 frames. ], batch size: 20, lr: 7.64e-03, grad_scale: 8.0 +2023-02-06 09:00:12,900 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1972, 3.1426, 2.8980, 1.4861, 2.8223, 2.7855, 2.8948, 2.6511], + device='cuda:1'), covar=tensor([0.1317, 0.0891, 0.1411, 0.4889, 0.1146, 0.1330, 0.1619, 0.1069], + device='cuda:1'), in_proj_covar=tensor([0.0453, 0.0357, 0.0375, 0.0471, 0.0369, 0.0358, 0.0369, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 09:00:20,728 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9328, 1.6666, 1.7065, 1.5049, 1.1265, 1.5918, 2.2131, 1.8762], + device='cuda:1'), covar=tensor([0.0450, 0.1203, 0.1717, 0.1471, 0.0606, 0.1492, 0.0655, 0.0634], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0155, 0.0195, 0.0161, 0.0106, 0.0165, 0.0119, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 09:00:25,447 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.953e+02 3.048e+02 3.591e+02 4.434e+02 8.168e+02, threshold=7.182e+02, percent-clipped=8.0 +2023-02-06 09:00:27,558 INFO [train.py:901] (1/4) Epoch 10, batch 4700, loss[loss=0.2657, simple_loss=0.3268, pruned_loss=0.1023, over 8804.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3139, pruned_loss=0.08378, over 1606551.91 frames. ], batch size: 40, lr: 7.64e-03, grad_scale: 8.0 +2023-02-06 09:00:33,883 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77457.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:00:58,287 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4640, 2.7295, 1.8657, 2.2343, 1.9977, 1.4707, 1.8849, 2.2059], + device='cuda:1'), covar=tensor([0.1253, 0.0315, 0.1007, 0.0511, 0.0703, 0.1457, 0.0914, 0.0797], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0235, 0.0313, 0.0297, 0.0301, 0.0325, 0.0340, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 09:01:02,744 INFO [train.py:901] (1/4) Epoch 10, batch 4750, loss[loss=0.2303, simple_loss=0.2904, pruned_loss=0.08513, over 7539.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3143, pruned_loss=0.08436, over 1605813.46 frames. ], batch size: 18, lr: 7.64e-03, grad_scale: 8.0 +2023-02-06 09:01:28,523 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 09:01:30,525 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 09:01:35,879 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.648e+02 3.312e+02 4.103e+02 1.054e+03, threshold=6.623e+02, percent-clipped=5.0 +2023-02-06 09:01:37,936 INFO [train.py:901] (1/4) Epoch 10, batch 4800, loss[loss=0.2602, simple_loss=0.333, pruned_loss=0.09371, over 8293.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3163, pruned_loss=0.08515, over 1612531.32 frames. ], batch size: 23, lr: 7.64e-03, grad_scale: 8.0 +2023-02-06 09:01:54,069 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77572.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:01:54,110 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77572.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:01:54,628 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:10,895 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77597.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:11,381 INFO [train.py:901] (1/4) Epoch 10, batch 4850, loss[loss=0.2452, simple_loss=0.3126, pruned_loss=0.08892, over 8032.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3168, pruned_loss=0.08516, over 1617064.70 frames. ], batch size: 22, lr: 7.63e-03, grad_scale: 8.0 +2023-02-06 09:02:16,292 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 09:02:38,004 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:38,637 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:42,019 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77639.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:46,049 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.616e+02 3.128e+02 3.870e+02 7.279e+02, threshold=6.256e+02, percent-clipped=1.0 +2023-02-06 09:02:48,055 INFO [train.py:901] (1/4) Epoch 10, batch 4900, loss[loss=0.2522, simple_loss=0.3218, pruned_loss=0.09128, over 8698.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3164, pruned_loss=0.08509, over 1609763.54 frames. ], batch size: 34, lr: 7.63e-03, grad_scale: 8.0 +2023-02-06 09:02:55,137 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1917, 1.3989, 4.1512, 1.8239, 2.4376, 4.7647, 4.7273, 4.0603], + device='cuda:1'), covar=tensor([0.1048, 0.1668, 0.0305, 0.1900, 0.1050, 0.0176, 0.0424, 0.0620], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0286, 0.0248, 0.0279, 0.0261, 0.0228, 0.0317, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 09:03:09,345 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4483, 1.4441, 2.8804, 1.2216, 2.0536, 3.0707, 3.1234, 2.5944], + device='cuda:1'), covar=tensor([0.1144, 0.1409, 0.0395, 0.2071, 0.0836, 0.0277, 0.0539, 0.0715], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0287, 0.0247, 0.0278, 0.0260, 0.0227, 0.0316, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 09:03:15,422 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77688.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:03:19,988 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77695.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:03:21,730 INFO [train.py:901] (1/4) Epoch 10, batch 4950, loss[loss=0.2749, simple_loss=0.3302, pruned_loss=0.1098, over 7544.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3163, pruned_loss=0.08575, over 1605776.16 frames. ], batch size: 18, lr: 7.63e-03, grad_scale: 8.0 +2023-02-06 09:03:54,524 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.777e+02 2.775e+02 3.348e+02 4.012e+02 9.680e+02, threshold=6.695e+02, percent-clipped=4.0 +2023-02-06 09:03:57,188 INFO [train.py:901] (1/4) Epoch 10, batch 5000, loss[loss=0.2767, simple_loss=0.3526, pruned_loss=0.1004, over 8515.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3179, pruned_loss=0.0869, over 1603694.93 frames. ], batch size: 28, lr: 7.63e-03, grad_scale: 8.0 +2023-02-06 09:03:58,711 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:03:59,989 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5018, 2.8595, 1.9533, 2.1811, 2.2049, 1.5125, 2.0121, 2.1385], + device='cuda:1'), covar=tensor([0.1686, 0.0317, 0.1034, 0.0740, 0.0704, 0.1502, 0.1109, 0.1116], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0234, 0.0312, 0.0299, 0.0305, 0.0326, 0.0339, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 09:04:01,945 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:04:30,717 INFO [train.py:901] (1/4) Epoch 10, batch 5050, loss[loss=0.254, simple_loss=0.3321, pruned_loss=0.08797, over 8470.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3178, pruned_loss=0.08654, over 1608322.19 frames. ], batch size: 25, lr: 7.62e-03, grad_scale: 8.0 +2023-02-06 09:04:51,127 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77828.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:04:52,883 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 09:05:02,965 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.592e+02 3.051e+02 4.098e+02 9.089e+02, threshold=6.102e+02, percent-clipped=4.0 +2023-02-06 09:05:05,650 INFO [train.py:901] (1/4) Epoch 10, batch 5100, loss[loss=0.3114, simple_loss=0.3561, pruned_loss=0.1333, over 6921.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3175, pruned_loss=0.08646, over 1610379.88 frames. ], batch size: 72, lr: 7.62e-03, grad_scale: 16.0 +2023-02-06 09:05:09,145 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77853.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:05:40,071 INFO [train.py:901] (1/4) Epoch 10, batch 5150, loss[loss=0.2116, simple_loss=0.2982, pruned_loss=0.06245, over 7963.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3165, pruned_loss=0.08542, over 1616054.92 frames. ], batch size: 21, lr: 7.62e-03, grad_scale: 16.0 +2023-02-06 09:05:43,179 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-02-06 09:06:07,894 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9007, 2.2278, 4.7380, 1.3270, 3.4207, 2.2717, 1.7514, 2.8800], + device='cuda:1'), covar=tensor([0.1500, 0.2130, 0.0606, 0.3746, 0.1189, 0.2557, 0.1674, 0.2095], + device='cuda:1'), in_proj_covar=tensor([0.0482, 0.0502, 0.0526, 0.0567, 0.0604, 0.0545, 0.0460, 0.0600], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 09:06:11,279 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:06:11,738 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.968e+02 2.805e+02 3.349e+02 3.898e+02 8.134e+02, threshold=6.697e+02, percent-clipped=4.0 +2023-02-06 09:06:13,812 INFO [train.py:901] (1/4) Epoch 10, batch 5200, loss[loss=0.2606, simple_loss=0.3295, pruned_loss=0.09583, over 8451.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3151, pruned_loss=0.08434, over 1616517.72 frames. ], batch size: 49, lr: 7.62e-03, grad_scale: 16.0 +2023-02-06 09:06:29,782 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:06:35,818 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77978.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:06:50,976 INFO [train.py:901] (1/4) Epoch 10, batch 5250, loss[loss=0.2201, simple_loss=0.2938, pruned_loss=0.07316, over 7814.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3139, pruned_loss=0.0836, over 1618096.86 frames. ], batch size: 20, lr: 7.61e-03, grad_scale: 16.0 +2023-02-06 09:06:56,860 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 09:06:57,815 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:07:00,644 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78010.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:07:05,372 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7955, 2.2204, 3.6884, 2.6465, 3.0695, 2.3647, 1.9312, 1.8502], + device='cuda:1'), covar=tensor([0.3383, 0.4225, 0.0980, 0.2869, 0.2090, 0.2133, 0.1611, 0.4405], + device='cuda:1'), in_proj_covar=tensor([0.0871, 0.0846, 0.0710, 0.0823, 0.0919, 0.0778, 0.0694, 0.0752], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:07:14,856 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78031.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:07:17,402 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78035.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:07:19,992 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78039.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:07:23,912 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.672e+02 3.378e+02 4.041e+02 9.848e+02, threshold=6.756e+02, percent-clipped=3.0 +2023-02-06 09:07:25,981 INFO [train.py:901] (1/4) Epoch 10, batch 5300, loss[loss=0.2642, simple_loss=0.3383, pruned_loss=0.09505, over 8293.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3131, pruned_loss=0.08295, over 1617488.96 frames. ], batch size: 23, lr: 7.61e-03, grad_scale: 16.0 +2023-02-06 09:07:57,639 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78093.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:08:00,803 INFO [train.py:901] (1/4) Epoch 10, batch 5350, loss[loss=0.2432, simple_loss=0.3183, pruned_loss=0.0841, over 8020.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3135, pruned_loss=0.08286, over 1618589.27 frames. ], batch size: 22, lr: 7.61e-03, grad_scale: 16.0 +2023-02-06 09:08:06,683 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2248, 1.8862, 2.7919, 2.3119, 2.5372, 2.1442, 1.6797, 1.2645], + device='cuda:1'), covar=tensor([0.3732, 0.3754, 0.1075, 0.2232, 0.1655, 0.2118, 0.1798, 0.3953], + device='cuda:1'), in_proj_covar=tensor([0.0869, 0.0846, 0.0708, 0.0819, 0.0916, 0.0777, 0.0692, 0.0748], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:08:34,365 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.713e+02 3.238e+02 4.266e+02 6.892e+02, threshold=6.476e+02, percent-clipped=1.0 +2023-02-06 09:08:35,741 INFO [train.py:901] (1/4) Epoch 10, batch 5400, loss[loss=0.2475, simple_loss=0.3243, pruned_loss=0.08534, over 8519.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3147, pruned_loss=0.08352, over 1622575.76 frames. ], batch size: 39, lr: 7.61e-03, grad_scale: 8.0 +2023-02-06 09:08:40,007 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78154.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:09:08,707 INFO [train.py:901] (1/4) Epoch 10, batch 5450, loss[loss=0.2176, simple_loss=0.2982, pruned_loss=0.06846, over 8451.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3145, pruned_loss=0.08361, over 1619606.35 frames. ], batch size: 27, lr: 7.60e-03, grad_scale: 8.0 +2023-02-06 09:09:43,437 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.697e+02 3.396e+02 4.413e+02 8.943e+02, threshold=6.791e+02, percent-clipped=7.0 +2023-02-06 09:09:44,799 INFO [train.py:901] (1/4) Epoch 10, batch 5500, loss[loss=0.1749, simple_loss=0.2548, pruned_loss=0.04748, over 7543.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3148, pruned_loss=0.08379, over 1615951.22 frames. ], batch size: 18, lr: 7.60e-03, grad_scale: 8.0 +2023-02-06 09:09:46,331 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:09:46,835 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 09:10:03,188 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6510, 2.4714, 4.5706, 1.3395, 3.2117, 2.2493, 1.7868, 2.7962], + device='cuda:1'), covar=tensor([0.1696, 0.2029, 0.0760, 0.4055, 0.1412, 0.2692, 0.1776, 0.2255], + device='cuda:1'), in_proj_covar=tensor([0.0482, 0.0501, 0.0524, 0.0570, 0.0602, 0.0544, 0.0460, 0.0599], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 09:10:17,484 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3214, 1.2137, 1.4649, 1.1363, 0.7703, 1.2845, 1.1458, 1.1437], + device='cuda:1'), covar=tensor([0.0527, 0.1357, 0.1692, 0.1478, 0.0555, 0.1601, 0.0739, 0.0633], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0155, 0.0194, 0.0160, 0.0105, 0.0164, 0.0119, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 09:10:18,554 INFO [train.py:901] (1/4) Epoch 10, batch 5550, loss[loss=0.247, simple_loss=0.3245, pruned_loss=0.08478, over 8431.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.3141, pruned_loss=0.08311, over 1615948.56 frames. ], batch size: 27, lr: 7.60e-03, grad_scale: 8.0 +2023-02-06 09:10:22,039 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4371, 2.9265, 1.7720, 2.1664, 2.0541, 1.5119, 1.8437, 2.2938], + device='cuda:1'), covar=tensor([0.1409, 0.0361, 0.1083, 0.0647, 0.0760, 0.1547, 0.1167, 0.0813], + device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0231, 0.0310, 0.0297, 0.0303, 0.0323, 0.0337, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 09:10:52,568 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.601e+02 2.563e+02 3.102e+02 4.076e+02 7.679e+02, threshold=6.204e+02, percent-clipped=2.0 +2023-02-06 09:10:54,680 INFO [train.py:901] (1/4) Epoch 10, batch 5600, loss[loss=0.2202, simple_loss=0.2963, pruned_loss=0.07203, over 7661.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3137, pruned_loss=0.08307, over 1614094.91 frames. ], batch size: 19, lr: 7.60e-03, grad_scale: 8.0 +2023-02-06 09:10:55,582 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:11:12,346 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78374.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:11:28,352 INFO [train.py:901] (1/4) Epoch 10, batch 5650, loss[loss=0.1995, simple_loss=0.2773, pruned_loss=0.06081, over 7255.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3137, pruned_loss=0.08311, over 1611127.73 frames. ], batch size: 16, lr: 7.59e-03, grad_scale: 8.0 +2023-02-06 09:11:36,903 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78410.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:11:48,304 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 09:11:54,461 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:12:02,394 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.443e+02 2.913e+02 3.480e+02 5.594e+02, threshold=5.826e+02, percent-clipped=0.0 +2023-02-06 09:12:03,751 INFO [train.py:901] (1/4) Epoch 10, batch 5700, loss[loss=0.2733, simple_loss=0.3342, pruned_loss=0.1062, over 8086.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3145, pruned_loss=0.08335, over 1616000.66 frames. ], batch size: 21, lr: 7.59e-03, grad_scale: 8.0 +2023-02-06 09:12:25,607 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:12:36,575 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2463, 2.4693, 1.9673, 2.9253, 1.2488, 1.5411, 1.9074, 2.2827], + device='cuda:1'), covar=tensor([0.0694, 0.0782, 0.1108, 0.0352, 0.1276, 0.1514, 0.1062, 0.0866], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0209, 0.0253, 0.0214, 0.0215, 0.0250, 0.0255, 0.0218], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 09:12:39,152 INFO [train.py:901] (1/4) Epoch 10, batch 5750, loss[loss=0.1927, simple_loss=0.2577, pruned_loss=0.06386, over 7531.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.3137, pruned_loss=0.08323, over 1612512.65 frames. ], batch size: 18, lr: 7.59e-03, grad_scale: 8.0 +2023-02-06 09:12:53,267 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 09:13:11,252 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 2.793e+02 3.478e+02 4.404e+02 1.244e+03, threshold=6.955e+02, percent-clipped=11.0 +2023-02-06 09:13:12,614 INFO [train.py:901] (1/4) Epoch 10, batch 5800, loss[loss=0.2191, simple_loss=0.296, pruned_loss=0.07109, over 8252.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3119, pruned_loss=0.08249, over 1609047.87 frames. ], batch size: 22, lr: 7.59e-03, grad_scale: 8.0 +2023-02-06 09:13:31,641 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78574.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:13:45,424 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:13:48,025 INFO [train.py:901] (1/4) Epoch 10, batch 5850, loss[loss=0.2289, simple_loss=0.2995, pruned_loss=0.07914, over 8015.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3128, pruned_loss=0.08268, over 1611555.06 frames. ], batch size: 22, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:14:12,162 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7060, 1.9304, 2.1595, 1.4819, 2.2640, 1.5837, 0.9293, 1.7408], + device='cuda:1'), covar=tensor([0.0392, 0.0191, 0.0124, 0.0320, 0.0218, 0.0475, 0.0516, 0.0202], + device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0308, 0.0265, 0.0376, 0.0305, 0.0462, 0.0352, 0.0337], + device='cuda:1'), out_proj_covar=tensor([1.1013e-04, 8.7315e-05, 7.5397e-05, 1.0727e-04, 8.8303e-05, 1.4328e-04, + 1.0199e-04, 9.6771e-05], device='cuda:1') +2023-02-06 09:14:19,904 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.589e+02 3.164e+02 4.281e+02 9.296e+02, threshold=6.329e+02, percent-clipped=5.0 +2023-02-06 09:14:21,265 INFO [train.py:901] (1/4) Epoch 10, batch 5900, loss[loss=0.2146, simple_loss=0.2989, pruned_loss=0.06521, over 8099.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3124, pruned_loss=0.08275, over 1610390.83 frames. ], batch size: 23, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:14:22,174 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7696, 1.4834, 1.9708, 1.6193, 1.8266, 1.6753, 1.3910, 0.7172], + device='cuda:1'), covar=tensor([0.3231, 0.2993, 0.1023, 0.1874, 0.1407, 0.1827, 0.1487, 0.2955], + device='cuda:1'), in_proj_covar=tensor([0.0872, 0.0849, 0.0710, 0.0827, 0.0917, 0.0776, 0.0692, 0.0745], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:14:47,571 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9594, 1.6556, 1.7603, 1.5701, 1.0717, 1.7647, 2.2422, 1.9962], + device='cuda:1'), covar=tensor([0.0431, 0.1205, 0.1667, 0.1327, 0.0600, 0.1392, 0.0610, 0.0557], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0155, 0.0195, 0.0159, 0.0106, 0.0165, 0.0118, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 09:14:57,566 INFO [train.py:901] (1/4) Epoch 10, batch 5950, loss[loss=0.2411, simple_loss=0.3191, pruned_loss=0.08158, over 8442.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3128, pruned_loss=0.08295, over 1611699.36 frames. ], batch size: 27, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:15:05,423 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:15:30,046 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.430e+02 2.939e+02 3.954e+02 7.661e+02, threshold=5.878e+02, percent-clipped=3.0 +2023-02-06 09:15:30,239 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3124, 1.3090, 2.2865, 1.2172, 2.2239, 2.4766, 2.5228, 2.1018], + device='cuda:1'), covar=tensor([0.0941, 0.1161, 0.0465, 0.1742, 0.0543, 0.0373, 0.0555, 0.0771], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0293, 0.0254, 0.0285, 0.0267, 0.0232, 0.0323, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 09:15:31,440 INFO [train.py:901] (1/4) Epoch 10, batch 6000, loss[loss=0.1998, simple_loss=0.2657, pruned_loss=0.06689, over 6365.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3125, pruned_loss=0.08325, over 1604084.05 frames. ], batch size: 14, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:15:31,441 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 09:15:43,951 INFO [train.py:935] (1/4) Epoch 10, validation: loss=0.1914, simple_loss=0.2907, pruned_loss=0.04604, over 944034.00 frames. +2023-02-06 09:15:43,952 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 09:16:18,420 INFO [train.py:901] (1/4) Epoch 10, batch 6050, loss[loss=0.2383, simple_loss=0.3011, pruned_loss=0.08775, over 7253.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3109, pruned_loss=0.08255, over 1595892.35 frames. ], batch size: 16, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:16:35,967 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78822.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:16:52,866 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.842e+02 3.348e+02 4.641e+02 9.072e+02, threshold=6.696e+02, percent-clipped=15.0 +2023-02-06 09:16:54,182 INFO [train.py:901] (1/4) Epoch 10, batch 6100, loss[loss=0.2141, simple_loss=0.2794, pruned_loss=0.07438, over 7698.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3113, pruned_loss=0.08241, over 1598126.58 frames. ], batch size: 18, lr: 7.57e-03, grad_scale: 8.0 +2023-02-06 09:17:24,370 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 09:17:27,730 INFO [train.py:901] (1/4) Epoch 10, batch 6150, loss[loss=0.2244, simple_loss=0.3027, pruned_loss=0.07302, over 7419.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3098, pruned_loss=0.08136, over 1599435.86 frames. ], batch size: 17, lr: 7.57e-03, grad_scale: 8.0 +2023-02-06 09:17:41,304 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78918.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:17:54,630 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78937.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:18:01,037 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.486e+02 3.076e+02 3.632e+02 7.166e+02, threshold=6.152e+02, percent-clipped=1.0 +2023-02-06 09:18:02,465 INFO [train.py:901] (1/4) Epoch 10, batch 6200, loss[loss=0.2361, simple_loss=0.308, pruned_loss=0.08214, over 7437.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3117, pruned_loss=0.08169, over 1606760.68 frames. ], batch size: 17, lr: 7.57e-03, grad_scale: 8.0 +2023-02-06 09:18:15,663 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:18:28,058 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78983.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:18:32,912 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78990.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:18:38,064 INFO [train.py:901] (1/4) Epoch 10, batch 6250, loss[loss=0.2387, simple_loss=0.3154, pruned_loss=0.08097, over 8360.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3116, pruned_loss=0.08212, over 1606400.28 frames. ], batch size: 24, lr: 7.57e-03, grad_scale: 8.0 +2023-02-06 09:19:01,794 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79033.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:19:10,145 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.509e+02 3.177e+02 4.128e+02 1.006e+03, threshold=6.355e+02, percent-clipped=7.0 +2023-02-06 09:19:11,558 INFO [train.py:901] (1/4) Epoch 10, batch 6300, loss[loss=0.2433, simple_loss=0.3208, pruned_loss=0.08287, over 8040.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3112, pruned_loss=0.08188, over 1599922.51 frames. ], batch size: 22, lr: 7.56e-03, grad_scale: 8.0 +2023-02-06 09:19:47,655 INFO [train.py:901] (1/4) Epoch 10, batch 6350, loss[loss=0.305, simple_loss=0.3412, pruned_loss=0.1345, over 7549.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3113, pruned_loss=0.08226, over 1599455.42 frames. ], batch size: 18, lr: 7.56e-03, grad_scale: 8.0 +2023-02-06 09:20:20,619 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.998e+02 3.636e+02 4.667e+02 1.201e+03, threshold=7.271e+02, percent-clipped=11.0 +2023-02-06 09:20:21,298 INFO [train.py:901] (1/4) Epoch 10, batch 6400, loss[loss=0.2399, simple_loss=0.3125, pruned_loss=0.08367, over 8473.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3139, pruned_loss=0.08387, over 1606650.01 frames. ], batch size: 25, lr: 7.56e-03, grad_scale: 8.0 +2023-02-06 09:20:54,163 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79193.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:20:57,437 INFO [train.py:901] (1/4) Epoch 10, batch 6450, loss[loss=0.2502, simple_loss=0.3268, pruned_loss=0.08676, over 8765.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.313, pruned_loss=0.08257, over 1610042.41 frames. ], batch size: 30, lr: 7.56e-03, grad_scale: 8.0 +2023-02-06 09:21:12,224 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79218.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:21:31,632 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.577e+02 3.130e+02 4.050e+02 7.383e+02, threshold=6.260e+02, percent-clipped=1.0 +2023-02-06 09:21:32,338 INFO [train.py:901] (1/4) Epoch 10, batch 6500, loss[loss=0.2126, simple_loss=0.2946, pruned_loss=0.06527, over 8204.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.313, pruned_loss=0.08226, over 1613606.81 frames. ], batch size: 23, lr: 7.55e-03, grad_scale: 8.0 +2023-02-06 09:21:37,198 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5702, 1.9430, 2.0759, 1.3026, 2.2276, 1.3909, 0.7276, 1.7820], + device='cuda:1'), covar=tensor([0.0380, 0.0203, 0.0151, 0.0300, 0.0212, 0.0590, 0.0475, 0.0183], + device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0308, 0.0262, 0.0374, 0.0298, 0.0460, 0.0349, 0.0334], + device='cuda:1'), out_proj_covar=tensor([1.0847e-04, 8.7017e-05, 7.4642e-05, 1.0684e-04, 8.5972e-05, 1.4267e-04, + 1.0098e-04, 9.6060e-05], device='cuda:1') +2023-02-06 09:21:59,879 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79289.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:22:06,996 INFO [train.py:901] (1/4) Epoch 10, batch 6550, loss[loss=0.2506, simple_loss=0.3268, pruned_loss=0.08726, over 8111.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3116, pruned_loss=0.08153, over 1612206.64 frames. ], batch size: 23, lr: 7.55e-03, grad_scale: 8.0 +2023-02-06 09:22:11,145 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79303.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:22:18,364 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79314.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:22:27,871 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:22:35,417 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4722, 1.7947, 2.7453, 1.2355, 1.9358, 1.8683, 1.5524, 1.7626], + device='cuda:1'), covar=tensor([0.1706, 0.1975, 0.0764, 0.3767, 0.1539, 0.2690, 0.1787, 0.2141], + device='cuda:1'), in_proj_covar=tensor([0.0482, 0.0502, 0.0527, 0.0569, 0.0612, 0.0545, 0.0463, 0.0603], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 09:22:36,078 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5069, 1.8618, 1.9141, 1.0896, 2.0989, 1.4281, 0.4755, 1.7974], + device='cuda:1'), covar=tensor([0.0329, 0.0214, 0.0163, 0.0284, 0.0201, 0.0556, 0.0496, 0.0143], + device='cuda:1'), in_proj_covar=tensor([0.0379, 0.0310, 0.0266, 0.0377, 0.0301, 0.0464, 0.0353, 0.0337], + device='cuda:1'), out_proj_covar=tensor([1.0953e-04, 8.7883e-05, 7.5710e-05, 1.0731e-04, 8.6790e-05, 1.4378e-04, + 1.0216e-04, 9.6991e-05], device='cuda:1') +2023-02-06 09:22:36,562 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 09:22:41,242 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.767e+02 3.312e+02 4.239e+02 1.073e+03, threshold=6.623e+02, percent-clipped=3.0 +2023-02-06 09:22:41,949 INFO [train.py:901] (1/4) Epoch 10, batch 6600, loss[loss=0.2494, simple_loss=0.3243, pruned_loss=0.08719, over 8660.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3121, pruned_loss=0.08202, over 1612557.36 frames. ], batch size: 39, lr: 7.55e-03, grad_scale: 8.0 +2023-02-06 09:22:44,151 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79351.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:22:53,887 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 09:23:04,704 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79382.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:23:11,890 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79393.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 09:23:15,197 INFO [train.py:901] (1/4) Epoch 10, batch 6650, loss[loss=0.243, simple_loss=0.2982, pruned_loss=0.09396, over 7206.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3134, pruned_loss=0.08299, over 1609133.05 frames. ], batch size: 16, lr: 7.55e-03, grad_scale: 8.0 +2023-02-06 09:23:15,338 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.4864, 1.5233, 5.5471, 2.1146, 4.9659, 4.6933, 5.1164, 4.9771], + device='cuda:1'), covar=tensor([0.0378, 0.4078, 0.0375, 0.3076, 0.0864, 0.0673, 0.0379, 0.0467], + device='cuda:1'), in_proj_covar=tensor([0.0469, 0.0564, 0.0570, 0.0521, 0.0596, 0.0505, 0.0496, 0.0571], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:23:47,673 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79442.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:23:47,686 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8991, 2.0830, 1.7575, 2.6425, 1.1568, 1.4613, 1.8378, 2.0980], + device='cuda:1'), covar=tensor([0.0792, 0.0929, 0.1084, 0.0413, 0.1184, 0.1572, 0.0882, 0.0774], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0212, 0.0254, 0.0216, 0.0218, 0.0252, 0.0258, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 09:23:50,857 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 2.666e+02 3.220e+02 4.193e+02 8.839e+02, threshold=6.440e+02, percent-clipped=3.0 +2023-02-06 09:23:51,582 INFO [train.py:901] (1/4) Epoch 10, batch 6700, loss[loss=0.2382, simple_loss=0.322, pruned_loss=0.07722, over 8198.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3148, pruned_loss=0.08357, over 1613485.52 frames. ], batch size: 23, lr: 7.54e-03, grad_scale: 8.0 +2023-02-06 09:24:11,621 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.5790, 1.8485, 3.7080, 1.9674, 3.3287, 3.1652, 3.4258, 3.3390], + device='cuda:1'), covar=tensor([0.0601, 0.2951, 0.0782, 0.2863, 0.0960, 0.0872, 0.0530, 0.0627], + device='cuda:1'), in_proj_covar=tensor([0.0469, 0.0565, 0.0572, 0.0523, 0.0595, 0.0507, 0.0497, 0.0569], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:24:24,671 INFO [train.py:901] (1/4) Epoch 10, batch 6750, loss[loss=0.2491, simple_loss=0.3224, pruned_loss=0.08795, over 8342.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3154, pruned_loss=0.08435, over 1611585.88 frames. ], batch size: 26, lr: 7.54e-03, grad_scale: 8.0 +2023-02-06 09:25:00,368 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.601e+02 2.662e+02 3.188e+02 4.113e+02 8.575e+02, threshold=6.376e+02, percent-clipped=4.0 +2023-02-06 09:25:01,059 INFO [train.py:901] (1/4) Epoch 10, batch 6800, loss[loss=0.218, simple_loss=0.3014, pruned_loss=0.06735, over 8362.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.315, pruned_loss=0.08398, over 1614244.25 frames. ], batch size: 24, lr: 7.54e-03, grad_scale: 8.0 +2023-02-06 09:25:11,666 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 09:25:35,820 INFO [train.py:901] (1/4) Epoch 10, batch 6850, loss[loss=0.2654, simple_loss=0.344, pruned_loss=0.09339, over 8480.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3141, pruned_loss=0.08384, over 1608390.76 frames. ], batch size: 29, lr: 7.54e-03, grad_scale: 8.0 +2023-02-06 09:25:39,355 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:25:59,652 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 09:26:10,492 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.480e+02 2.958e+02 3.519e+02 6.592e+02, threshold=5.916e+02, percent-clipped=1.0 +2023-02-06 09:26:10,580 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79647.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 09:26:11,121 INFO [train.py:901] (1/4) Epoch 10, batch 6900, loss[loss=0.2396, simple_loss=0.3236, pruned_loss=0.07779, over 8026.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3133, pruned_loss=0.08324, over 1607423.82 frames. ], batch size: 22, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:26:19,350 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 09:26:26,448 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0543, 1.2632, 1.2021, 0.6510, 1.2670, 1.0317, 0.2231, 1.1466], + device='cuda:1'), covar=tensor([0.0228, 0.0211, 0.0197, 0.0314, 0.0250, 0.0570, 0.0479, 0.0186], + device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0305, 0.0261, 0.0370, 0.0296, 0.0456, 0.0349, 0.0333], + device='cuda:1'), out_proj_covar=tensor([1.0844e-04, 8.6142e-05, 7.4195e-05, 1.0539e-04, 8.5244e-05, 1.4116e-04, + 1.0118e-04, 9.5813e-05], device='cuda:1') +2023-02-06 09:26:30,985 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79675.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:26:44,404 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79695.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:26:46,401 INFO [train.py:901] (1/4) Epoch 10, batch 6950, loss[loss=0.2084, simple_loss=0.2996, pruned_loss=0.05863, over 8473.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3124, pruned_loss=0.08195, over 1614465.73 frames. ], batch size: 25, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:26:46,615 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:27:01,842 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 09:27:03,563 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79723.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:27:05,484 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79726.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:27:10,661 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 09:27:12,697 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79737.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 09:27:19,259 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.682e+02 2.770e+02 3.379e+02 4.019e+02 1.115e+03, threshold=6.759e+02, percent-clipped=8.0 +2023-02-06 09:27:19,985 INFO [train.py:901] (1/4) Epoch 10, batch 7000, loss[loss=0.1963, simple_loss=0.2676, pruned_loss=0.0625, over 7433.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3122, pruned_loss=0.08198, over 1611912.63 frames. ], batch size: 17, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:27:28,518 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 09:27:30,359 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79762.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:27:43,454 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79780.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:27:55,338 INFO [train.py:901] (1/4) Epoch 10, batch 7050, loss[loss=0.2691, simple_loss=0.3411, pruned_loss=0.09857, over 8288.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3119, pruned_loss=0.08166, over 1611955.07 frames. ], batch size: 23, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:27:58,325 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2525, 1.2920, 2.3242, 1.1858, 2.0829, 2.4711, 2.5497, 2.1388], + device='cuda:1'), covar=tensor([0.1078, 0.1300, 0.0489, 0.2033, 0.0724, 0.0461, 0.0677, 0.0829], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0293, 0.0251, 0.0284, 0.0268, 0.0233, 0.0328, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 09:28:04,455 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:28:18,695 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2424, 1.5271, 1.6615, 1.3872, 1.0044, 1.5254, 1.7571, 1.8529], + device='cuda:1'), covar=tensor([0.0507, 0.1274, 0.1708, 0.1414, 0.0651, 0.1556, 0.0735, 0.0537], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0152, 0.0193, 0.0159, 0.0105, 0.0164, 0.0117, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:1') +2023-02-06 09:28:25,535 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:28:29,363 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.704e+02 3.361e+02 4.306e+02 1.362e+03, threshold=6.722e+02, percent-clipped=5.0 +2023-02-06 09:28:29,796 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 09:28:30,079 INFO [train.py:901] (1/4) Epoch 10, batch 7100, loss[loss=0.2511, simple_loss=0.3307, pruned_loss=0.0857, over 8454.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3123, pruned_loss=0.08194, over 1612627.95 frames. ], batch size: 27, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:28:32,891 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79852.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:29:06,028 INFO [train.py:901] (1/4) Epoch 10, batch 7150, loss[loss=0.2321, simple_loss=0.3031, pruned_loss=0.08057, over 8182.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3127, pruned_loss=0.08159, over 1616037.01 frames. ], batch size: 23, lr: 7.52e-03, grad_scale: 8.0 +2023-02-06 09:29:33,297 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.91 vs. limit=5.0 +2023-02-06 09:29:39,477 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.702e+02 3.262e+02 4.332e+02 1.613e+03, threshold=6.525e+02, percent-clipped=3.0 +2023-02-06 09:29:39,575 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79947.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:29:40,184 INFO [train.py:901] (1/4) Epoch 10, batch 7200, loss[loss=0.2482, simple_loss=0.3311, pruned_loss=0.08268, over 8546.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3119, pruned_loss=0.08108, over 1611119.72 frames. ], batch size: 31, lr: 7.52e-03, grad_scale: 8.0 +2023-02-06 09:30:13,866 INFO [train.py:901] (1/4) Epoch 10, batch 7250, loss[loss=0.2315, simple_loss=0.3067, pruned_loss=0.07813, over 8315.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3124, pruned_loss=0.08187, over 1611405.41 frames. ], batch size: 25, lr: 7.52e-03, grad_scale: 8.0 +2023-02-06 09:30:30,522 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80018.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 09:30:31,018 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80019.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:30:47,885 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80043.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:30:50,296 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.755e+02 3.243e+02 3.993e+02 1.489e+03, threshold=6.485e+02, percent-clipped=9.0 +2023-02-06 09:30:50,951 INFO [train.py:901] (1/4) Epoch 10, batch 7300, loss[loss=0.2545, simple_loss=0.3392, pruned_loss=0.08491, over 8357.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3135, pruned_loss=0.08301, over 1608164.51 frames. ], batch size: 26, lr: 7.52e-03, grad_scale: 8.0 +2023-02-06 09:31:00,594 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:03,354 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:04,718 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8171, 2.0163, 1.6861, 2.6049, 1.0968, 1.3991, 1.6874, 2.1667], + device='cuda:1'), covar=tensor([0.0871, 0.0946, 0.1111, 0.0455, 0.1285, 0.1554, 0.1109, 0.0711], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0212, 0.0253, 0.0216, 0.0218, 0.0250, 0.0257, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 09:31:20,006 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80091.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:24,178 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80097.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:24,600 INFO [train.py:901] (1/4) Epoch 10, batch 7350, loss[loss=0.291, simple_loss=0.3602, pruned_loss=0.1109, over 8331.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.3136, pruned_loss=0.08356, over 1609470.84 frames. ], batch size: 25, lr: 7.51e-03, grad_scale: 8.0 +2023-02-06 09:31:31,679 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80108.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 09:31:42,410 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80122.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:44,304 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80124.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:50,388 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80133.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:31:50,994 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:56,728 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 09:31:59,434 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.519e+02 3.343e+02 4.224e+02 9.659e+02, threshold=6.686e+02, percent-clipped=6.0 +2023-02-06 09:32:00,144 INFO [train.py:901] (1/4) Epoch 10, batch 7400, loss[loss=0.1911, simple_loss=0.2731, pruned_loss=0.05456, over 7925.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3148, pruned_loss=0.08442, over 1609701.25 frames. ], batch size: 20, lr: 7.51e-03, grad_scale: 8.0 +2023-02-06 09:32:16,199 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 09:32:19,815 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7267, 2.2963, 4.2788, 1.4425, 2.9739, 2.1371, 1.8331, 2.5435], + device='cuda:1'), covar=tensor([0.1834, 0.2221, 0.0722, 0.4179, 0.1768, 0.3129, 0.1823, 0.2618], + device='cuda:1'), in_proj_covar=tensor([0.0490, 0.0512, 0.0536, 0.0576, 0.0617, 0.0558, 0.0468, 0.0608], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 09:32:34,304 INFO [train.py:901] (1/4) Epoch 10, batch 7450, loss[loss=0.2311, simple_loss=0.3101, pruned_loss=0.07601, over 8360.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3154, pruned_loss=0.08424, over 1614351.95 frames. ], batch size: 24, lr: 7.51e-03, grad_scale: 8.0 +2023-02-06 09:32:34,500 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9448, 1.6301, 1.7126, 1.5316, 1.1566, 1.6022, 1.7042, 1.6235], + device='cuda:1'), covar=tensor([0.0544, 0.0995, 0.1375, 0.1109, 0.0601, 0.1211, 0.0685, 0.0474], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0153, 0.0195, 0.0159, 0.0105, 0.0164, 0.0118, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 09:32:47,421 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 09:32:54,358 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 09:33:02,478 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80239.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:33:07,266 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9172, 1.0386, 3.0072, 0.9209, 2.6432, 2.5221, 2.7967, 2.6921], + device='cuda:1'), covar=tensor([0.0632, 0.3438, 0.0692, 0.2884, 0.1254, 0.0916, 0.0597, 0.0722], + device='cuda:1'), in_proj_covar=tensor([0.0460, 0.0554, 0.0558, 0.0508, 0.0582, 0.0490, 0.0484, 0.0551], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:33:09,144 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.637e+02 3.217e+02 3.901e+02 6.824e+02, threshold=6.433e+02, percent-clipped=2.0 +2023-02-06 09:33:09,864 INFO [train.py:901] (1/4) Epoch 10, batch 7500, loss[loss=0.3119, simple_loss=0.3697, pruned_loss=0.1271, over 8525.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3148, pruned_loss=0.08396, over 1611351.33 frames. ], batch size: 26, lr: 7.51e-03, grad_scale: 8.0 +2023-02-06 09:33:43,946 INFO [train.py:901] (1/4) Epoch 10, batch 7550, loss[loss=0.273, simple_loss=0.3515, pruned_loss=0.0972, over 8450.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3134, pruned_loss=0.08261, over 1613890.42 frames. ], batch size: 27, lr: 7.50e-03, grad_scale: 8.0 +2023-02-06 09:33:46,245 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80301.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:33:57,962 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80318.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:34:07,069 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 +2023-02-06 09:34:07,862 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.87 vs. limit=5.0 +2023-02-06 09:34:15,033 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:34:17,454 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.399e+02 2.908e+02 3.933e+02 1.078e+03, threshold=5.816e+02, percent-clipped=3.0 +2023-02-06 09:34:18,142 INFO [train.py:901] (1/4) Epoch 10, batch 7600, loss[loss=0.2789, simple_loss=0.3624, pruned_loss=0.09773, over 8698.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3149, pruned_loss=0.0832, over 1613610.42 frames. ], batch size: 49, lr: 7.50e-03, grad_scale: 8.0 +2023-02-06 09:34:32,353 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2145, 1.1345, 4.4199, 1.9935, 2.6610, 5.1054, 5.0292, 4.4180], + device='cuda:1'), covar=tensor([0.1072, 0.1999, 0.0298, 0.1845, 0.1003, 0.0182, 0.0409, 0.0580], + device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0294, 0.0253, 0.0286, 0.0268, 0.0231, 0.0329, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 09:34:49,202 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80390.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:34:54,346 INFO [train.py:901] (1/4) Epoch 10, batch 7650, loss[loss=0.2542, simple_loss=0.3237, pruned_loss=0.09234, over 8709.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3132, pruned_loss=0.08263, over 1611701.54 frames. ], batch size: 39, lr: 7.50e-03, grad_scale: 8.0 +2023-02-06 09:35:03,331 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80411.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:35:05,493 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4468, 1.9532, 2.9847, 2.2783, 2.5714, 2.2699, 1.8387, 1.2129], + device='cuda:1'), covar=tensor([0.3533, 0.3820, 0.1117, 0.2684, 0.2025, 0.2083, 0.1662, 0.4089], + device='cuda:1'), in_proj_covar=tensor([0.0882, 0.0856, 0.0723, 0.0830, 0.0930, 0.0790, 0.0700, 0.0761], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:35:06,117 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80415.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:35:10,826 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4408, 1.8015, 1.8955, 1.0442, 1.9913, 1.4749, 0.3929, 1.5613], + device='cuda:1'), covar=tensor([0.0336, 0.0200, 0.0186, 0.0298, 0.0280, 0.0518, 0.0494, 0.0150], + device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0305, 0.0262, 0.0374, 0.0295, 0.0453, 0.0345, 0.0337], + device='cuda:1'), out_proj_covar=tensor([1.0765e-04, 8.6215e-05, 7.4130e-05, 1.0662e-04, 8.4942e-05, 1.3962e-04, + 9.9863e-05, 9.6723e-05], device='cuda:1') +2023-02-06 09:35:27,278 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.621e+02 3.149e+02 3.913e+02 9.838e+02, threshold=6.298e+02, percent-clipped=6.0 +2023-02-06 09:35:27,989 INFO [train.py:901] (1/4) Epoch 10, batch 7700, loss[loss=0.2322, simple_loss=0.3155, pruned_loss=0.07448, over 8465.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3123, pruned_loss=0.08229, over 1607431.46 frames. ], batch size: 25, lr: 7.50e-03, grad_scale: 8.0 +2023-02-06 09:35:51,538 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80481.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:35:54,393 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.60 vs. limit=5.0 +2023-02-06 09:36:01,536 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80495.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:36:01,919 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 09:36:03,344 INFO [train.py:901] (1/4) Epoch 10, batch 7750, loss[loss=0.2259, simple_loss=0.292, pruned_loss=0.07992, over 7803.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3118, pruned_loss=0.08217, over 1608168.78 frames. ], batch size: 19, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:36:06,754 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 09:36:13,282 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80512.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:36:13,358 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7723, 2.0517, 2.3399, 1.4553, 2.4041, 1.7242, 0.8655, 1.9241], + device='cuda:1'), covar=tensor([0.0406, 0.0227, 0.0172, 0.0354, 0.0249, 0.0579, 0.0516, 0.0186], + device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0303, 0.0261, 0.0373, 0.0294, 0.0451, 0.0344, 0.0336], + device='cuda:1'), out_proj_covar=tensor([1.0755e-04, 8.5373e-05, 7.3923e-05, 1.0623e-04, 8.4408e-05, 1.3903e-04, + 9.9296e-05, 9.6469e-05], device='cuda:1') +2023-02-06 09:36:18,623 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:36:36,346 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.713e+02 3.406e+02 4.090e+02 8.759e+02, threshold=6.812e+02, percent-clipped=3.0 +2023-02-06 09:36:37,056 INFO [train.py:901] (1/4) Epoch 10, batch 7800, loss[loss=0.2211, simple_loss=0.3032, pruned_loss=0.06949, over 8108.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3127, pruned_loss=0.08286, over 1606624.47 frames. ], batch size: 23, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:36:49,780 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7860, 3.8309, 3.3955, 1.9117, 3.3112, 3.3666, 3.4429, 3.0611], + device='cuda:1'), covar=tensor([0.0888, 0.0600, 0.1074, 0.4431, 0.1028, 0.0988, 0.1333, 0.0947], + device='cuda:1'), in_proj_covar=tensor([0.0460, 0.0357, 0.0379, 0.0477, 0.0377, 0.0363, 0.0367, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 09:37:09,857 INFO [train.py:901] (1/4) Epoch 10, batch 7850, loss[loss=0.2685, simple_loss=0.3463, pruned_loss=0.09538, over 8456.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3123, pruned_loss=0.08244, over 1604100.03 frames. ], batch size: 27, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:37:40,874 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:37:42,726 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.556e+02 3.372e+02 4.255e+02 7.191e+02, threshold=6.744e+02, percent-clipped=1.0 +2023-02-06 09:37:43,427 INFO [train.py:901] (1/4) Epoch 10, batch 7900, loss[loss=0.2597, simple_loss=0.3332, pruned_loss=0.09309, over 8438.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3126, pruned_loss=0.08297, over 1605857.82 frames. ], batch size: 27, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:37:56,467 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 09:37:57,225 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.39 vs. limit=5.0 +2023-02-06 09:38:11,818 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9810, 1.6651, 1.9144, 1.5933, 1.1284, 1.9649, 2.1485, 2.0715], + device='cuda:1'), covar=tensor([0.0457, 0.1203, 0.1628, 0.1337, 0.0597, 0.1368, 0.0670, 0.0551], + device='cuda:1'), in_proj_covar=tensor([0.0103, 0.0155, 0.0195, 0.0159, 0.0106, 0.0165, 0.0119, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 09:38:16,929 INFO [train.py:901] (1/4) Epoch 10, batch 7950, loss[loss=0.2277, simple_loss=0.2906, pruned_loss=0.08247, over 7657.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3121, pruned_loss=0.08245, over 1610742.29 frames. ], batch size: 19, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:38:50,732 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.443e+02 2.660e+02 3.023e+02 3.700e+02 9.606e+02, threshold=6.046e+02, percent-clipped=2.0 +2023-02-06 09:38:51,446 INFO [train.py:901] (1/4) Epoch 10, batch 8000, loss[loss=0.2407, simple_loss=0.314, pruned_loss=0.08369, over 8476.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3129, pruned_loss=0.08301, over 1607401.70 frames. ], batch size: 49, lr: 7.48e-03, grad_scale: 8.0 +2023-02-06 09:38:55,690 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:38:56,306 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:38:59,808 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:39:14,662 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 09:39:20,370 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80791.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:39:24,988 INFO [train.py:901] (1/4) Epoch 10, batch 8050, loss[loss=0.2204, simple_loss=0.2921, pruned_loss=0.07439, over 7285.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.3133, pruned_loss=0.0837, over 1602007.09 frames. ], batch size: 16, lr: 7.48e-03, grad_scale: 8.0 +2023-02-06 09:39:43,567 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80825.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:39:58,221 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 09:40:01,794 INFO [train.py:901] (1/4) Epoch 11, batch 0, loss[loss=0.2893, simple_loss=0.3533, pruned_loss=0.1127, over 8524.00 frames. ], tot_loss[loss=0.2893, simple_loss=0.3533, pruned_loss=0.1127, over 8524.00 frames. ], batch size: 49, lr: 7.14e-03, grad_scale: 8.0 +2023-02-06 09:40:01,795 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 09:40:13,090 INFO [train.py:935] (1/4) Epoch 11, validation: loss=0.1907, simple_loss=0.2907, pruned_loss=0.04534, over 944034.00 frames. +2023-02-06 09:40:13,091 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 09:40:23,915 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.799e+02 2.899e+02 3.439e+02 4.416e+02 1.589e+03, threshold=6.879e+02, percent-clipped=9.0 +2023-02-06 09:40:27,457 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 09:40:30,151 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:40:39,870 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80870.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:40:47,942 INFO [train.py:901] (1/4) Epoch 11, batch 50, loss[loss=0.2359, simple_loss=0.3145, pruned_loss=0.0787, over 8194.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3181, pruned_loss=0.08411, over 367693.05 frames. ], batch size: 23, lr: 7.14e-03, grad_scale: 8.0 +2023-02-06 09:40:54,280 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5229, 1.3901, 2.7726, 1.2380, 1.9830, 3.0658, 3.1065, 2.5847], + device='cuda:1'), covar=tensor([0.1182, 0.1552, 0.0449, 0.2082, 0.0933, 0.0296, 0.0546, 0.0678], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0291, 0.0252, 0.0283, 0.0268, 0.0231, 0.0326, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 09:40:57,788 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3251, 1.3276, 2.3172, 1.2070, 2.1140, 2.4940, 2.5864, 2.1336], + device='cuda:1'), covar=tensor([0.0995, 0.1231, 0.0460, 0.1976, 0.0685, 0.0387, 0.0600, 0.0706], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0291, 0.0251, 0.0282, 0.0268, 0.0230, 0.0325, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 09:41:03,890 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 09:41:15,730 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7622, 2.6398, 1.9979, 4.3515, 1.6179, 1.6683, 2.7622, 2.8882], + device='cuda:1'), covar=tensor([0.1786, 0.1447, 0.2028, 0.0201, 0.1738, 0.2410, 0.1154, 0.1181], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0217, 0.0259, 0.0220, 0.0222, 0.0259, 0.0259, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 09:41:24,350 INFO [train.py:901] (1/4) Epoch 11, batch 100, loss[loss=0.2287, simple_loss=0.2816, pruned_loss=0.08792, over 7447.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3137, pruned_loss=0.08177, over 640371.14 frames. ], batch size: 17, lr: 7.14e-03, grad_scale: 8.0 +2023-02-06 09:41:29,239 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 09:41:30,713 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80940.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:41:35,314 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.679e+02 3.187e+02 3.933e+02 1.063e+03, threshold=6.374e+02, percent-clipped=2.0 +2023-02-06 09:41:50,783 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 09:41:51,851 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80971.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:41:58,397 INFO [train.py:901] (1/4) Epoch 11, batch 150, loss[loss=0.2875, simple_loss=0.362, pruned_loss=0.1065, over 8569.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3142, pruned_loss=0.0823, over 858359.17 frames. ], batch size: 39, lr: 7.13e-03, grad_scale: 8.0 +2023-02-06 09:42:23,729 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81016.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:42:34,633 INFO [train.py:901] (1/4) Epoch 11, batch 200, loss[loss=0.219, simple_loss=0.3007, pruned_loss=0.06868, over 8230.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3156, pruned_loss=0.08408, over 1025395.43 frames. ], batch size: 22, lr: 7.13e-03, grad_scale: 8.0 +2023-02-06 09:42:42,993 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81041.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:42:47,013 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.935e+02 2.662e+02 3.186e+02 4.005e+02 8.686e+02, threshold=6.371e+02, percent-clipped=5.0 +2023-02-06 09:43:10,553 INFO [train.py:901] (1/4) Epoch 11, batch 250, loss[loss=0.2153, simple_loss=0.3008, pruned_loss=0.06491, over 8608.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.3136, pruned_loss=0.08297, over 1156483.34 frames. ], batch size: 49, lr: 7.13e-03, grad_scale: 8.0 +2023-02-06 09:43:21,546 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 09:43:22,296 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81098.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:43:31,348 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 09:43:36,941 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2307, 1.5506, 1.6582, 1.4329, 1.0623, 1.5418, 1.9468, 1.6614], + device='cuda:1'), covar=tensor([0.0470, 0.1255, 0.1691, 0.1404, 0.0616, 0.1539, 0.0658, 0.0603], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0154, 0.0195, 0.0159, 0.0107, 0.0164, 0.0118, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 09:43:42,713 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81126.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:43:46,009 INFO [train.py:901] (1/4) Epoch 11, batch 300, loss[loss=0.2432, simple_loss=0.3004, pruned_loss=0.093, over 7815.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3116, pruned_loss=0.08174, over 1258573.19 frames. ], batch size: 19, lr: 7.13e-03, grad_scale: 16.0 +2023-02-06 09:43:48,260 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:43:48,875 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:43:57,127 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.697e+02 3.136e+02 4.054e+02 9.565e+02, threshold=6.271e+02, percent-clipped=1.0 +2023-02-06 09:44:00,775 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81151.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:22,514 INFO [train.py:901] (1/4) Epoch 11, batch 350, loss[loss=0.2394, simple_loss=0.3091, pruned_loss=0.08489, over 7990.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.312, pruned_loss=0.08158, over 1340853.97 frames. ], batch size: 21, lr: 7.13e-03, grad_scale: 8.0 +2023-02-06 09:44:32,494 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-06 09:44:33,016 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81196.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:39,210 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0227, 1.6506, 2.2307, 1.8805, 2.0472, 1.9721, 1.6957, 0.6676], + device='cuda:1'), covar=tensor([0.4000, 0.3603, 0.1187, 0.2110, 0.1603, 0.1966, 0.1597, 0.3637], + device='cuda:1'), in_proj_covar=tensor([0.0892, 0.0863, 0.0720, 0.0833, 0.0935, 0.0794, 0.0704, 0.0762], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:44:44,357 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:46,355 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9233, 1.7479, 1.8620, 1.6974, 1.2489, 1.7366, 2.2949, 1.7517], + device='cuda:1'), covar=tensor([0.0452, 0.1172, 0.1599, 0.1290, 0.0564, 0.1390, 0.0630, 0.0606], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0155, 0.0195, 0.0159, 0.0106, 0.0164, 0.0118, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 09:44:49,728 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81221.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:53,755 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81227.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:56,281 INFO [train.py:901] (1/4) Epoch 11, batch 400, loss[loss=0.2629, simple_loss=0.3363, pruned_loss=0.09475, over 8702.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3128, pruned_loss=0.08223, over 1401447.65 frames. ], batch size: 34, lr: 7.12e-03, grad_scale: 8.0 +2023-02-06 09:45:00,801 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 +2023-02-06 09:45:08,651 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.601e+02 3.216e+02 4.274e+02 6.931e+02, threshold=6.433e+02, percent-clipped=2.0 +2023-02-06 09:45:10,287 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:45:11,747 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81252.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:45:23,643 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4845, 1.7567, 1.8952, 1.0800, 1.9494, 1.2748, 0.4021, 1.6752], + device='cuda:1'), covar=tensor([0.0375, 0.0220, 0.0165, 0.0365, 0.0228, 0.0689, 0.0565, 0.0167], + device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0304, 0.0259, 0.0372, 0.0293, 0.0457, 0.0345, 0.0339], + device='cuda:1'), out_proj_covar=tensor([1.0753e-04, 8.5703e-05, 7.3022e-05, 1.0561e-04, 8.3944e-05, 1.4109e-04, + 9.9503e-05, 9.6983e-05], device='cuda:1') +2023-02-06 09:45:29,085 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9147, 3.8123, 2.6992, 2.5997, 2.9691, 2.1601, 3.0239, 2.9284], + device='cuda:1'), covar=tensor([0.1394, 0.0286, 0.0767, 0.0631, 0.0511, 0.1102, 0.0774, 0.0887], + device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0233, 0.0308, 0.0296, 0.0303, 0.0325, 0.0338, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 09:45:32,911 INFO [train.py:901] (1/4) Epoch 11, batch 450, loss[loss=0.218, simple_loss=0.3014, pruned_loss=0.06733, over 8361.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3134, pruned_loss=0.08195, over 1453868.09 frames. ], batch size: 24, lr: 7.12e-03, grad_scale: 8.0 +2023-02-06 09:45:57,089 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81317.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:45:58,459 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81319.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:46:06,253 INFO [train.py:901] (1/4) Epoch 11, batch 500, loss[loss=0.2103, simple_loss=0.2917, pruned_loss=0.06446, over 8028.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3126, pruned_loss=0.08117, over 1491791.37 frames. ], batch size: 22, lr: 7.12e-03, grad_scale: 8.0 +2023-02-06 09:46:07,904 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.83 vs. limit=5.0 +2023-02-06 09:46:17,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.501e+02 3.364e+02 4.069e+02 6.845e+02, threshold=6.728e+02, percent-clipped=2.0 +2023-02-06 09:46:40,091 INFO [train.py:901] (1/4) Epoch 11, batch 550, loss[loss=0.1891, simple_loss=0.2736, pruned_loss=0.05234, over 7807.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3119, pruned_loss=0.08091, over 1519246.41 frames. ], batch size: 20, lr: 7.12e-03, grad_scale: 8.0 +2023-02-06 09:47:04,541 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2552, 1.9004, 2.7393, 2.1705, 2.5520, 2.1107, 1.7646, 1.1479], + device='cuda:1'), covar=tensor([0.3963, 0.3748, 0.1134, 0.2384, 0.1694, 0.2228, 0.1772, 0.3862], + device='cuda:1'), in_proj_covar=tensor([0.0886, 0.0853, 0.0717, 0.0828, 0.0927, 0.0789, 0.0701, 0.0759], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:47:15,699 INFO [train.py:901] (1/4) Epoch 11, batch 600, loss[loss=0.273, simple_loss=0.3448, pruned_loss=0.1006, over 7810.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3119, pruned_loss=0.08121, over 1539222.46 frames. ], batch size: 20, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:47:27,353 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.440e+02 2.633e+02 3.080e+02 3.885e+02 6.931e+02, threshold=6.160e+02, percent-clipped=1.0 +2023-02-06 09:47:27,526 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8182, 1.5674, 3.1882, 1.2624, 2.1869, 3.4426, 3.4650, 2.9070], + device='cuda:1'), covar=tensor([0.1089, 0.1409, 0.0343, 0.1965, 0.0795, 0.0255, 0.0519, 0.0651], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0286, 0.0247, 0.0278, 0.0264, 0.0226, 0.0322, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-02-06 09:47:27,548 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8616, 1.3725, 1.6694, 1.2448, 1.0814, 1.4952, 1.7583, 1.6098], + device='cuda:1'), covar=tensor([0.0523, 0.1326, 0.1747, 0.1460, 0.0612, 0.1515, 0.0703, 0.0600], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0155, 0.0196, 0.0159, 0.0106, 0.0165, 0.0118, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 09:47:27,967 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 09:47:35,471 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 09:47:42,475 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81469.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:47:48,472 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:47:50,444 INFO [train.py:901] (1/4) Epoch 11, batch 650, loss[loss=0.2225, simple_loss=0.306, pruned_loss=0.06945, over 8034.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3124, pruned_loss=0.08151, over 1559219.54 frames. ], batch size: 22, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:47:59,446 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81494.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:48:04,899 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1510, 2.2276, 1.6970, 2.0414, 1.7511, 1.4367, 1.7466, 1.7138], + device='cuda:1'), covar=tensor([0.1119, 0.0330, 0.0944, 0.0410, 0.0585, 0.1292, 0.0788, 0.0671], + device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0228, 0.0304, 0.0295, 0.0297, 0.0318, 0.0332, 0.0300], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 09:48:08,337 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81506.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:48:26,933 INFO [train.py:901] (1/4) Epoch 11, batch 700, loss[loss=0.2602, simple_loss=0.3353, pruned_loss=0.09261, over 8700.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3109, pruned_loss=0.08078, over 1569934.06 frames. ], batch size: 40, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:48:27,091 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:48:28,472 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0167, 2.2168, 3.5948, 1.5530, 2.9339, 2.3271, 2.0955, 2.6112], + device='cuda:1'), covar=tensor([0.1455, 0.2292, 0.0630, 0.3844, 0.1414, 0.2472, 0.1586, 0.2290], + device='cuda:1'), in_proj_covar=tensor([0.0480, 0.0504, 0.0526, 0.0567, 0.0609, 0.0541, 0.0466, 0.0608], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 09:48:38,762 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.541e+02 3.049e+02 3.626e+02 6.264e+02, threshold=6.097e+02, percent-clipped=1.0 +2023-02-06 09:49:01,395 INFO [train.py:901] (1/4) Epoch 11, batch 750, loss[loss=0.2535, simple_loss=0.3196, pruned_loss=0.09364, over 7696.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3108, pruned_loss=0.08033, over 1579982.79 frames. ], batch size: 18, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:49:07,835 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-06 09:49:10,212 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81593.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:49:10,586 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 09:49:21,636 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3230, 1.8914, 3.0207, 2.2879, 2.6518, 2.1567, 1.6580, 1.3022], + device='cuda:1'), covar=tensor([0.3754, 0.4105, 0.1001, 0.2446, 0.1868, 0.2120, 0.1822, 0.3880], + device='cuda:1'), in_proj_covar=tensor([0.0885, 0.0855, 0.0718, 0.0833, 0.0931, 0.0790, 0.0701, 0.0761], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:49:24,850 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 09:49:25,673 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3427, 1.5293, 1.3708, 1.8170, 0.8024, 1.2413, 1.2887, 1.5210], + device='cuda:1'), covar=tensor([0.0914, 0.0768, 0.1122, 0.0524, 0.1142, 0.1421, 0.0856, 0.0730], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0212, 0.0254, 0.0216, 0.0216, 0.0254, 0.0257, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 09:49:29,019 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7743, 5.9901, 5.0311, 2.2676, 5.2147, 5.5389, 5.6156, 5.3263], + device='cuda:1'), covar=tensor([0.0621, 0.0436, 0.1039, 0.5184, 0.0736, 0.0638, 0.1050, 0.0687], + device='cuda:1'), in_proj_covar=tensor([0.0465, 0.0368, 0.0379, 0.0479, 0.0373, 0.0365, 0.0367, 0.0322], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 09:49:33,722 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 09:49:36,455 INFO [train.py:901] (1/4) Epoch 11, batch 800, loss[loss=0.2375, simple_loss=0.3008, pruned_loss=0.08707, over 7811.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3098, pruned_loss=0.08019, over 1585505.67 frames. ], batch size: 20, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:49:49,270 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.494e+02 2.971e+02 3.970e+02 9.403e+02, threshold=5.941e+02, percent-clipped=2.0 +2023-02-06 09:49:58,265 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:49:59,646 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81663.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:50:11,555 INFO [train.py:901] (1/4) Epoch 11, batch 850, loss[loss=0.2205, simple_loss=0.2917, pruned_loss=0.07466, over 7228.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3114, pruned_loss=0.08056, over 1594570.83 frames. ], batch size: 16, lr: 7.10e-03, grad_scale: 8.0 +2023-02-06 09:50:16,503 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81688.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:50:46,088 INFO [train.py:901] (1/4) Epoch 11, batch 900, loss[loss=0.2466, simple_loss=0.3352, pruned_loss=0.07898, over 8525.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3115, pruned_loss=0.08043, over 1603397.27 frames. ], batch size: 29, lr: 7.10e-03, grad_scale: 8.0 +2023-02-06 09:50:58,849 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 2.543e+02 3.289e+02 4.286e+02 9.063e+02, threshold=6.577e+02, percent-clipped=7.0 +2023-02-06 09:51:08,752 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0399, 1.4591, 1.4615, 1.2408, 1.1149, 1.2979, 1.7382, 1.3554], + device='cuda:1'), covar=tensor([0.0541, 0.1280, 0.1791, 0.1492, 0.0593, 0.1595, 0.0734, 0.0655], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0154, 0.0195, 0.0159, 0.0106, 0.0165, 0.0118, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 09:51:18,680 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81776.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:51:20,092 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81778.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:51:21,851 INFO [train.py:901] (1/4) Epoch 11, batch 950, loss[loss=0.2174, simple_loss=0.2832, pruned_loss=0.07582, over 7807.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3125, pruned_loss=0.08159, over 1609703.83 frames. ], batch size: 20, lr: 7.10e-03, grad_scale: 8.0 +2023-02-06 09:51:37,583 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1026, 0.9926, 1.0983, 1.0944, 0.7641, 1.2036, 0.0395, 0.8939], + device='cuda:1'), covar=tensor([0.2452, 0.1638, 0.0610, 0.1105, 0.3663, 0.0633, 0.3311, 0.1653], + device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0164, 0.0094, 0.0210, 0.0253, 0.0102, 0.0162, 0.0159], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 09:51:51,854 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 09:51:56,016 INFO [train.py:901] (1/4) Epoch 11, batch 1000, loss[loss=0.2415, simple_loss=0.3151, pruned_loss=0.0839, over 7914.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3124, pruned_loss=0.08126, over 1609464.27 frames. ], batch size: 20, lr: 7.10e-03, grad_scale: 8.0 +2023-02-06 09:52:06,321 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4982, 1.8680, 3.1050, 1.2928, 2.2701, 1.8697, 1.5412, 2.1231], + device='cuda:1'), covar=tensor([0.1686, 0.2073, 0.0716, 0.3904, 0.1464, 0.2774, 0.1816, 0.2027], + device='cuda:1'), in_proj_covar=tensor([0.0485, 0.0515, 0.0537, 0.0582, 0.0618, 0.0552, 0.0474, 0.0614], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 09:52:07,440 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.713e+02 3.211e+02 4.023e+02 7.481e+02, threshold=6.422e+02, percent-clipped=3.0 +2023-02-06 09:52:08,391 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81849.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:52:13,880 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-06 09:52:22,680 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 09:52:27,198 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 09:52:27,395 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81874.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:52:31,957 INFO [train.py:901] (1/4) Epoch 11, batch 1050, loss[loss=0.1963, simple_loss=0.2783, pruned_loss=0.0571, over 7787.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3124, pruned_loss=0.08116, over 1610091.47 frames. ], batch size: 19, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:52:39,044 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 09:53:06,144 INFO [train.py:901] (1/4) Epoch 11, batch 1100, loss[loss=0.2332, simple_loss=0.3039, pruned_loss=0.08128, over 7984.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.312, pruned_loss=0.08112, over 1609724.27 frames. ], batch size: 21, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:53:18,507 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.436e+02 2.887e+02 3.709e+02 9.106e+02, threshold=5.774e+02, percent-clipped=2.0 +2023-02-06 09:53:36,920 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4062, 1.4508, 2.2847, 1.1372, 2.0264, 2.4994, 2.5474, 2.0925], + device='cuda:1'), covar=tensor([0.0936, 0.1098, 0.0477, 0.1955, 0.0666, 0.0367, 0.0582, 0.0759], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0297, 0.0255, 0.0287, 0.0271, 0.0232, 0.0332, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 09:53:36,945 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2231, 1.2579, 1.4408, 1.1859, 0.8587, 1.2055, 1.2918, 1.0817], + device='cuda:1'), covar=tensor([0.0542, 0.1220, 0.1667, 0.1320, 0.0544, 0.1525, 0.0660, 0.0611], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0153, 0.0193, 0.0157, 0.0104, 0.0164, 0.0117, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 09:53:41,609 INFO [train.py:901] (1/4) Epoch 11, batch 1150, loss[loss=0.2438, simple_loss=0.3136, pruned_loss=0.08702, over 8611.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3119, pruned_loss=0.08111, over 1612728.57 frames. ], batch size: 49, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:53:51,761 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 09:54:00,452 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:12,086 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2130, 1.8885, 2.7700, 2.2194, 2.5442, 2.1065, 1.6900, 1.2972], + device='cuda:1'), covar=tensor([0.3701, 0.3821, 0.1058, 0.2432, 0.1711, 0.2132, 0.1644, 0.3984], + device='cuda:1'), in_proj_covar=tensor([0.0869, 0.0840, 0.0710, 0.0819, 0.0907, 0.0778, 0.0690, 0.0749], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 09:54:17,961 INFO [train.py:901] (1/4) Epoch 11, batch 1200, loss[loss=0.1882, simple_loss=0.2575, pruned_loss=0.05952, over 7642.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3118, pruned_loss=0.08092, over 1615345.34 frames. ], batch size: 19, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:54:18,728 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:18,847 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:20,244 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82034.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:26,899 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3251, 1.8005, 3.0540, 1.3400, 2.1360, 3.3372, 3.3437, 2.8055], + device='cuda:1'), covar=tensor([0.0762, 0.1293, 0.0386, 0.2004, 0.0975, 0.0262, 0.0509, 0.0583], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0298, 0.0255, 0.0287, 0.0273, 0.0232, 0.0332, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 09:54:29,502 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.664e+02 3.172e+02 3.772e+02 1.117e+03, threshold=6.345e+02, percent-clipped=5.0 +2023-02-06 09:54:36,794 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:38,233 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:45,481 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-02-06 09:54:47,999 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82073.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:53,463 INFO [train.py:901] (1/4) Epoch 11, batch 1250, loss[loss=0.3083, simple_loss=0.3753, pruned_loss=0.1206, over 8097.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3117, pruned_loss=0.08073, over 1617425.33 frames. ], batch size: 23, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:55:29,355 INFO [train.py:901] (1/4) Epoch 11, batch 1300, loss[loss=0.2403, simple_loss=0.3181, pruned_loss=0.08129, over 8496.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3121, pruned_loss=0.08097, over 1617317.93 frames. ], batch size: 26, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:55:40,416 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82147.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:55:40,861 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.583e+02 3.223e+02 4.179e+02 7.623e+02, threshold=6.447e+02, percent-clipped=2.0 +2023-02-06 09:55:57,032 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1524, 1.0916, 1.2824, 1.0931, 0.8863, 1.3122, 0.2048, 0.8997], + device='cuda:1'), covar=tensor([0.2283, 0.1677, 0.0517, 0.1556, 0.3125, 0.0576, 0.3102, 0.1791], + device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0163, 0.0095, 0.0211, 0.0253, 0.0102, 0.0161, 0.0159], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 09:56:03,693 INFO [train.py:901] (1/4) Epoch 11, batch 1350, loss[loss=0.2432, simple_loss=0.3123, pruned_loss=0.08705, over 7648.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3141, pruned_loss=0.08256, over 1618825.57 frames. ], batch size: 19, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:56:38,846 INFO [train.py:901] (1/4) Epoch 11, batch 1400, loss[loss=0.276, simple_loss=0.3493, pruned_loss=0.1014, over 8249.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3136, pruned_loss=0.08224, over 1618592.80 frames. ], batch size: 24, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:56:51,063 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.627e+02 3.119e+02 3.954e+02 1.224e+03, threshold=6.238e+02, percent-clipped=1.0 +2023-02-06 09:56:55,576 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-02-06 09:57:13,605 INFO [train.py:901] (1/4) Epoch 11, batch 1450, loss[loss=0.2215, simple_loss=0.3019, pruned_loss=0.07059, over 7925.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3123, pruned_loss=0.08157, over 1615724.80 frames. ], batch size: 20, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:57:27,730 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 09:57:28,847 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.78 vs. limit=5.0 +2023-02-06 09:57:42,966 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 09:57:48,805 INFO [train.py:901] (1/4) Epoch 11, batch 1500, loss[loss=0.2821, simple_loss=0.3498, pruned_loss=0.1072, over 8696.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.3137, pruned_loss=0.08297, over 1608928.83 frames. ], batch size: 34, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:58:01,391 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.943e+02 2.743e+02 3.193e+02 4.270e+02 9.879e+02, threshold=6.387e+02, percent-clipped=7.0 +2023-02-06 09:58:02,230 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:13,191 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:16,175 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 09:58:24,635 INFO [train.py:901] (1/4) Epoch 11, batch 1550, loss[loss=0.2523, simple_loss=0.3258, pruned_loss=0.08943, over 8473.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3136, pruned_loss=0.08283, over 1612869.51 frames. ], batch size: 27, lr: 7.07e-03, grad_scale: 8.0 +2023-02-06 09:58:39,990 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82403.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:50,113 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:50,233 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3125, 2.9046, 2.3574, 4.0212, 1.6480, 2.1040, 2.2359, 3.0441], + device='cuda:1'), covar=tensor([0.0837, 0.0884, 0.0946, 0.0241, 0.1232, 0.1323, 0.1168, 0.0765], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0212, 0.0251, 0.0216, 0.0216, 0.0252, 0.0253, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 09:58:57,860 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82428.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:59,750 INFO [train.py:901] (1/4) Epoch 11, batch 1600, loss[loss=0.2306, simple_loss=0.3185, pruned_loss=0.07133, over 8100.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3137, pruned_loss=0.08199, over 1617236.35 frames. ], batch size: 23, lr: 7.07e-03, grad_scale: 8.0 +2023-02-06 09:59:12,998 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.328e+02 2.878e+02 3.468e+02 7.869e+02, threshold=5.757e+02, percent-clipped=2.0 +2023-02-06 09:59:24,134 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82464.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:59:36,407 INFO [train.py:901] (1/4) Epoch 11, batch 1650, loss[loss=0.2406, simple_loss=0.3216, pruned_loss=0.07983, over 8194.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3134, pruned_loss=0.0818, over 1618066.97 frames. ], batch size: 23, lr: 7.07e-03, grad_scale: 8.0 +2023-02-06 09:59:57,313 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82511.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:00:11,613 INFO [train.py:901] (1/4) Epoch 11, batch 1700, loss[loss=0.2552, simple_loss=0.3326, pruned_loss=0.08892, over 8529.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3138, pruned_loss=0.08227, over 1619571.48 frames. ], batch size: 49, lr: 7.07e-03, grad_scale: 8.0 +2023-02-06 10:00:12,454 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:00:17,650 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 10:00:23,206 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.517e+02 3.185e+02 4.066e+02 8.085e+02, threshold=6.370e+02, percent-clipped=5.0 +2023-02-06 10:00:47,528 INFO [train.py:901] (1/4) Epoch 11, batch 1750, loss[loss=0.238, simple_loss=0.3116, pruned_loss=0.08214, over 8499.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3128, pruned_loss=0.08198, over 1613068.97 frames. ], batch size: 26, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:00:52,548 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:01:23,306 INFO [train.py:901] (1/4) Epoch 11, batch 1800, loss[loss=0.2375, simple_loss=0.315, pruned_loss=0.07998, over 8469.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3117, pruned_loss=0.08141, over 1609251.41 frames. ], batch size: 25, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:01:35,823 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.598e+02 3.107e+02 4.193e+02 1.199e+03, threshold=6.213e+02, percent-clipped=8.0 +2023-02-06 10:01:58,592 INFO [train.py:901] (1/4) Epoch 11, batch 1850, loss[loss=0.2326, simple_loss=0.3041, pruned_loss=0.08053, over 7279.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3114, pruned_loss=0.08147, over 1607750.86 frames. ], batch size: 16, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:02:18,371 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82708.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:02:26,596 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82720.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:02:34,013 INFO [train.py:901] (1/4) Epoch 11, batch 1900, loss[loss=0.2494, simple_loss=0.3301, pruned_loss=0.08431, over 8339.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3107, pruned_loss=0.08082, over 1608626.97 frames. ], batch size: 26, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:02:43,694 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:02:45,528 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.672e+02 2.442e+02 3.142e+02 3.936e+02 6.780e+02, threshold=6.284e+02, percent-clipped=1.0 +2023-02-06 10:03:08,879 INFO [train.py:901] (1/4) Epoch 11, batch 1950, loss[loss=0.2322, simple_loss=0.3131, pruned_loss=0.07569, over 8479.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3107, pruned_loss=0.0809, over 1613752.00 frames. ], batch size: 25, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:03:12,343 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 10:03:13,853 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82788.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:03:26,471 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 10:03:32,231 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:03:39,051 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82823.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:03:44,652 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 10:03:45,035 INFO [train.py:901] (1/4) Epoch 11, batch 2000, loss[loss=0.173, simple_loss=0.2562, pruned_loss=0.04491, over 7536.00 frames. ], tot_loss[loss=0.236, simple_loss=0.3108, pruned_loss=0.08061, over 1615134.37 frames. ], batch size: 18, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:03:47,024 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 10:03:56,672 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.675e+02 3.279e+02 3.987e+02 1.082e+03, threshold=6.559e+02, percent-clipped=7.0 +2023-02-06 10:04:01,553 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82855.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:04:10,616 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0023, 1.2517, 1.1947, 0.5133, 1.2109, 1.0223, 0.0764, 1.1291], + device='cuda:1'), covar=tensor([0.0263, 0.0225, 0.0191, 0.0347, 0.0257, 0.0625, 0.0506, 0.0194], + device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0317, 0.0266, 0.0377, 0.0303, 0.0468, 0.0351, 0.0347], + device='cuda:1'), out_proj_covar=tensor([1.1015e-04, 8.9331e-05, 7.4800e-05, 1.0639e-04, 8.6493e-05, 1.4422e-04, + 1.0102e-04, 9.9094e-05], device='cuda:1') +2023-02-06 10:04:19,314 INFO [train.py:901] (1/4) Epoch 11, batch 2050, loss[loss=0.1898, simple_loss=0.2642, pruned_loss=0.05772, over 7237.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.31, pruned_loss=0.08022, over 1612537.85 frames. ], batch size: 16, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:04:55,307 INFO [train.py:901] (1/4) Epoch 11, batch 2100, loss[loss=0.1932, simple_loss=0.2788, pruned_loss=0.05381, over 8145.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3113, pruned_loss=0.08043, over 1618488.13 frames. ], batch size: 22, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:04:55,377 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82931.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:05:07,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.489e+02 3.174e+02 3.706e+02 9.083e+02, threshold=6.348e+02, percent-clipped=2.0 +2023-02-06 10:05:13,280 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8258, 1.6751, 1.8314, 1.6197, 0.9941, 1.7988, 2.3065, 1.9656], + device='cuda:1'), covar=tensor([0.0445, 0.1182, 0.1614, 0.1318, 0.0622, 0.1382, 0.0630, 0.0576], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0154, 0.0195, 0.0160, 0.0105, 0.0166, 0.0118, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 10:05:22,088 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82970.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:05:29,198 INFO [train.py:901] (1/4) Epoch 11, batch 2150, loss[loss=0.2346, simple_loss=0.2966, pruned_loss=0.08632, over 7790.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3121, pruned_loss=0.08078, over 1621549.72 frames. ], batch size: 19, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:05:29,371 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82981.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:05:49,414 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 10:06:04,068 INFO [train.py:901] (1/4) Epoch 11, batch 2200, loss[loss=0.2666, simple_loss=0.34, pruned_loss=0.09664, over 8109.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3116, pruned_loss=0.08069, over 1620549.32 frames. ], batch size: 23, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:06:15,769 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83046.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:06:16,953 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.510e+02 3.092e+02 4.104e+02 1.639e+03, threshold=6.185e+02, percent-clipped=4.0 +2023-02-06 10:06:38,935 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83079.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:06:40,017 INFO [train.py:901] (1/4) Epoch 11, batch 2250, loss[loss=0.1936, simple_loss=0.2721, pruned_loss=0.05755, over 7911.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3102, pruned_loss=0.08003, over 1616684.80 frames. ], batch size: 20, lr: 7.04e-03, grad_scale: 8.0 +2023-02-06 10:06:46,977 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0001, 1.5362, 1.3823, 1.6499, 1.3283, 1.2352, 1.3327, 1.3914], + device='cuda:1'), covar=tensor([0.1006, 0.0428, 0.1088, 0.0455, 0.0611, 0.1273, 0.0723, 0.0619], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0236, 0.0315, 0.0297, 0.0304, 0.0326, 0.0341, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 10:06:49,943 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 10:06:55,778 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83104.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:07:14,108 INFO [train.py:901] (1/4) Epoch 11, batch 2300, loss[loss=0.223, simple_loss=0.2959, pruned_loss=0.07504, over 8149.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3092, pruned_loss=0.08003, over 1611880.08 frames. ], batch size: 22, lr: 7.04e-03, grad_scale: 8.0 +2023-02-06 10:07:25,651 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.480e+02 3.199e+02 4.275e+02 9.806e+02, threshold=6.398e+02, percent-clipped=6.0 +2023-02-06 10:07:48,928 INFO [train.py:901] (1/4) Epoch 11, batch 2350, loss[loss=0.2179, simple_loss=0.308, pruned_loss=0.06391, over 8453.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3095, pruned_loss=0.0796, over 1616325.23 frames. ], batch size: 27, lr: 7.04e-03, grad_scale: 16.0 +2023-02-06 10:08:12,069 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83214.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:08:20,036 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83226.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:08:23,031 INFO [train.py:901] (1/4) Epoch 11, batch 2400, loss[loss=0.1942, simple_loss=0.2653, pruned_loss=0.06155, over 7267.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3096, pruned_loss=0.07993, over 1610735.43 frames. ], batch size: 16, lr: 7.04e-03, grad_scale: 16.0 +2023-02-06 10:08:35,100 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.359e+02 2.853e+02 3.666e+02 7.740e+02, threshold=5.706e+02, percent-clipped=1.0 +2023-02-06 10:08:37,261 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83251.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:08:58,651 INFO [train.py:901] (1/4) Epoch 11, batch 2450, loss[loss=0.2393, simple_loss=0.3123, pruned_loss=0.08313, over 8287.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.308, pruned_loss=0.07945, over 1611906.76 frames. ], batch size: 23, lr: 7.04e-03, grad_scale: 16.0 +2023-02-06 10:09:13,729 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83302.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:09:13,799 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83302.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:09:29,004 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83325.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:09:30,519 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:09:32,767 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 10:09:33,056 INFO [train.py:901] (1/4) Epoch 11, batch 2500, loss[loss=0.2909, simple_loss=0.3441, pruned_loss=0.1189, over 6897.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3093, pruned_loss=0.08042, over 1613897.79 frames. ], batch size: 72, lr: 7.03e-03, grad_scale: 16.0 +2023-02-06 10:09:44,609 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.719e+02 2.634e+02 3.143e+02 3.904e+02 7.323e+02, threshold=6.285e+02, percent-clipped=4.0 +2023-02-06 10:10:00,399 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1854, 1.8015, 2.5449, 2.0934, 2.3049, 2.0659, 1.7048, 1.0643], + device='cuda:1'), covar=tensor([0.3852, 0.3691, 0.1072, 0.2222, 0.1668, 0.1892, 0.1568, 0.3798], + device='cuda:1'), in_proj_covar=tensor([0.0878, 0.0859, 0.0727, 0.0832, 0.0932, 0.0785, 0.0700, 0.0758], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 10:10:07,376 INFO [train.py:901] (1/4) Epoch 11, batch 2550, loss[loss=0.2248, simple_loss=0.2998, pruned_loss=0.0749, over 7804.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3102, pruned_loss=0.08078, over 1612639.10 frames. ], batch size: 19, lr: 7.03e-03, grad_scale: 16.0 +2023-02-06 10:10:43,105 INFO [train.py:901] (1/4) Epoch 11, batch 2600, loss[loss=0.2126, simple_loss=0.2989, pruned_loss=0.06315, over 8289.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3101, pruned_loss=0.08016, over 1617423.57 frames. ], batch size: 23, lr: 7.03e-03, grad_scale: 16.0 +2023-02-06 10:10:49,495 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83440.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:10:54,713 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.898e+02 2.607e+02 3.192e+02 4.372e+02 8.439e+02, threshold=6.384e+02, percent-clipped=10.0 +2023-02-06 10:11:17,508 INFO [train.py:901] (1/4) Epoch 11, batch 2650, loss[loss=0.2358, simple_loss=0.311, pruned_loss=0.0803, over 8291.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3105, pruned_loss=0.08084, over 1617321.17 frames. ], batch size: 23, lr: 7.03e-03, grad_scale: 16.0 +2023-02-06 10:11:52,414 INFO [train.py:901] (1/4) Epoch 11, batch 2700, loss[loss=0.2216, simple_loss=0.3043, pruned_loss=0.06949, over 8651.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3103, pruned_loss=0.08093, over 1616503.99 frames. ], batch size: 27, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:12:04,667 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.578e+02 3.131e+02 4.095e+02 6.916e+02, threshold=6.263e+02, percent-clipped=2.0 +2023-02-06 10:12:06,306 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7281, 1.9177, 1.6423, 2.2955, 1.0914, 1.3893, 1.6033, 2.0033], + device='cuda:1'), covar=tensor([0.0833, 0.0869, 0.1116, 0.0496, 0.1181, 0.1608, 0.0961, 0.0779], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0215, 0.0252, 0.0217, 0.0217, 0.0254, 0.0257, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 10:12:11,630 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83558.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:12:17,332 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.11 vs. limit=5.0 +2023-02-06 10:12:27,325 INFO [train.py:901] (1/4) Epoch 11, batch 2750, loss[loss=0.215, simple_loss=0.3003, pruned_loss=0.0648, over 8282.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3104, pruned_loss=0.08075, over 1609795.65 frames. ], batch size: 23, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:12:49,818 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([6.1570, 1.7241, 6.2463, 2.0881, 5.7335, 5.3236, 5.8973, 5.7548], + device='cuda:1'), covar=tensor([0.0328, 0.3693, 0.0227, 0.2923, 0.0744, 0.0604, 0.0273, 0.0337], + device='cuda:1'), in_proj_covar=tensor([0.0478, 0.0567, 0.0577, 0.0524, 0.0596, 0.0498, 0.0501, 0.0571], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 10:12:51,398 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 10:13:03,333 INFO [train.py:901] (1/4) Epoch 11, batch 2800, loss[loss=0.1959, simple_loss=0.2807, pruned_loss=0.05553, over 8034.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3104, pruned_loss=0.08045, over 1611138.47 frames. ], batch size: 22, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:13:03,709 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 10:13:13,810 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83646.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:13:15,055 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.535e+02 3.136e+02 3.769e+02 1.201e+03, threshold=6.273e+02, percent-clipped=3.0 +2023-02-06 10:13:32,686 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83673.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:13:37,613 INFO [train.py:901] (1/4) Epoch 11, batch 2850, loss[loss=0.3307, simple_loss=0.3817, pruned_loss=0.1399, over 8505.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.309, pruned_loss=0.08022, over 1606325.80 frames. ], batch size: 28, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:13:47,918 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83696.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:13:49,189 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0474, 1.3974, 1.6166, 1.2829, 1.1070, 1.3904, 1.6864, 1.8337], + device='cuda:1'), covar=tensor([0.0510, 0.1230, 0.1780, 0.1398, 0.0571, 0.1491, 0.0694, 0.0545], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0154, 0.0194, 0.0159, 0.0105, 0.0164, 0.0117, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 10:14:05,650 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83721.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:14:12,966 INFO [train.py:901] (1/4) Epoch 11, batch 2900, loss[loss=0.2168, simple_loss=0.3008, pruned_loss=0.06647, over 8682.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3098, pruned_loss=0.08029, over 1607511.76 frames. ], batch size: 34, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:14:25,269 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.545e+02 3.159e+02 4.165e+02 9.643e+02, threshold=6.318e+02, percent-clipped=5.0 +2023-02-06 10:14:34,286 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:14:46,314 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83778.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:14:48,167 INFO [train.py:901] (1/4) Epoch 11, batch 2950, loss[loss=0.2421, simple_loss=0.3062, pruned_loss=0.08896, over 7262.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3108, pruned_loss=0.08096, over 1605897.02 frames. ], batch size: 16, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:14:53,616 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 10:15:09,067 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 10:15:18,756 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.78 vs. limit=5.0 +2023-02-06 10:15:22,299 INFO [train.py:901] (1/4) Epoch 11, batch 3000, loss[loss=0.2265, simple_loss=0.3024, pruned_loss=0.07531, over 7935.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3114, pruned_loss=0.08143, over 1606715.58 frames. ], batch size: 20, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:15:22,299 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 10:15:34,555 INFO [train.py:935] (1/4) Epoch 11, validation: loss=0.1889, simple_loss=0.2886, pruned_loss=0.04461, over 944034.00 frames. +2023-02-06 10:15:34,556 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 10:15:46,618 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.511e+02 2.977e+02 3.600e+02 5.313e+02, threshold=5.953e+02, percent-clipped=0.0 +2023-02-06 10:16:10,359 INFO [train.py:901] (1/4) Epoch 11, batch 3050, loss[loss=0.2228, simple_loss=0.3124, pruned_loss=0.06665, over 8353.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3122, pruned_loss=0.08217, over 1606028.46 frames. ], batch size: 24, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:16:21,308 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9178, 1.6139, 1.3382, 1.5861, 1.3772, 1.2009, 1.2330, 1.2817], + device='cuda:1'), covar=tensor([0.1042, 0.0409, 0.1146, 0.0518, 0.0593, 0.1279, 0.0878, 0.0691], + device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0234, 0.0314, 0.0295, 0.0299, 0.0320, 0.0338, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 10:16:43,140 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83929.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:16:44,272 INFO [train.py:901] (1/4) Epoch 11, batch 3100, loss[loss=0.1743, simple_loss=0.2598, pruned_loss=0.04442, over 7638.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3114, pruned_loss=0.08129, over 1607248.80 frames. ], batch size: 19, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:16:55,415 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.748e+02 3.262e+02 3.755e+02 7.942e+02, threshold=6.525e+02, percent-clipped=1.0 +2023-02-06 10:17:00,081 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83954.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:17:08,763 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83967.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:17:18,428 INFO [train.py:901] (1/4) Epoch 11, batch 3150, loss[loss=0.1815, simple_loss=0.2561, pruned_loss=0.05347, over 7545.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.311, pruned_loss=0.08113, over 1604901.10 frames. ], batch size: 18, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:17:34,199 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84003.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:17:42,318 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4384, 2.0993, 2.9183, 2.3589, 2.7236, 2.2800, 1.8584, 1.3032], + device='cuda:1'), covar=tensor([0.3728, 0.3786, 0.1116, 0.2630, 0.1979, 0.2108, 0.1595, 0.4362], + device='cuda:1'), in_proj_covar=tensor([0.0884, 0.0865, 0.0727, 0.0837, 0.0934, 0.0793, 0.0700, 0.0763], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 10:17:44,278 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84017.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:17:53,258 INFO [train.py:901] (1/4) Epoch 11, batch 3200, loss[loss=0.2384, simple_loss=0.3134, pruned_loss=0.08175, over 8429.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.312, pruned_loss=0.08162, over 1606849.63 frames. ], batch size: 49, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:18:01,393 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:18:05,770 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.726e+02 3.369e+02 4.220e+02 9.302e+02, threshold=6.739e+02, percent-clipped=4.0 +2023-02-06 10:18:27,189 INFO [train.py:901] (1/4) Epoch 11, batch 3250, loss[loss=0.2266, simple_loss=0.2997, pruned_loss=0.0768, over 7971.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3136, pruned_loss=0.08257, over 1613686.75 frames. ], batch size: 21, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:18:50,436 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84115.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:18:55,080 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84122.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:19:01,839 INFO [train.py:901] (1/4) Epoch 11, batch 3300, loss[loss=0.2656, simple_loss=0.3364, pruned_loss=0.09742, over 8688.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3134, pruned_loss=0.08208, over 1612428.70 frames. ], batch size: 34, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:19:08,603 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9138, 1.9720, 2.3623, 1.6045, 1.2073, 2.5569, 0.3991, 1.5633], + device='cuda:1'), covar=tensor([0.2161, 0.1635, 0.0520, 0.2230, 0.4322, 0.0422, 0.3355, 0.1778], + device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0167, 0.0097, 0.0215, 0.0253, 0.0104, 0.0166, 0.0165], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 10:19:13,378 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.547e+02 2.729e+02 3.101e+02 4.103e+02 8.191e+02, threshold=6.202e+02, percent-clipped=3.0 +2023-02-06 10:19:35,418 INFO [train.py:901] (1/4) Epoch 11, batch 3350, loss[loss=0.2216, simple_loss=0.3103, pruned_loss=0.06645, over 8103.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3148, pruned_loss=0.08331, over 1611969.81 frames. ], batch size: 23, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:19:37,882 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 10:20:01,020 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84217.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:20:10,209 INFO [train.py:901] (1/4) Epoch 11, batch 3400, loss[loss=0.1967, simple_loss=0.2856, pruned_loss=0.05391, over 8251.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3147, pruned_loss=0.08325, over 1615675.28 frames. ], batch size: 24, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:20:15,228 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84237.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:20:23,165 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 2.553e+02 3.068e+02 3.977e+02 7.727e+02, threshold=6.137e+02, percent-clipped=2.0 +2023-02-06 10:20:26,658 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84254.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:20:45,376 INFO [train.py:901] (1/4) Epoch 11, batch 3450, loss[loss=0.2279, simple_loss=0.3096, pruned_loss=0.07311, over 8450.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3137, pruned_loss=0.08234, over 1612259.61 frames. ], batch size: 27, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:21:06,407 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84311.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:21:20,277 INFO [train.py:901] (1/4) Epoch 11, batch 3500, loss[loss=0.234, simple_loss=0.3209, pruned_loss=0.07354, over 8202.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3133, pruned_loss=0.08202, over 1614527.06 frames. ], batch size: 23, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:21:31,076 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:21:32,267 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.703e+02 3.166e+02 4.187e+02 8.001e+02, threshold=6.332e+02, percent-clipped=6.0 +2023-02-06 10:21:36,449 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84354.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:21:48,766 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 10:21:51,923 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 10:21:54,137 INFO [train.py:901] (1/4) Epoch 11, batch 3550, loss[loss=0.2419, simple_loss=0.3231, pruned_loss=0.0803, over 8338.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3132, pruned_loss=0.08207, over 1613221.67 frames. ], batch size: 26, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:22:25,864 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84426.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:22:28,968 INFO [train.py:901] (1/4) Epoch 11, batch 3600, loss[loss=0.2521, simple_loss=0.3236, pruned_loss=0.09031, over 8081.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3143, pruned_loss=0.08313, over 1609302.00 frames. ], batch size: 21, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:22:37,194 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3131, 1.8997, 2.7705, 2.3393, 2.5758, 2.0906, 1.7068, 1.3412], + device='cuda:1'), covar=tensor([0.3764, 0.3744, 0.1238, 0.2360, 0.1853, 0.2132, 0.1604, 0.3831], + device='cuda:1'), in_proj_covar=tensor([0.0881, 0.0861, 0.0725, 0.0841, 0.0935, 0.0795, 0.0702, 0.0761], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 10:22:41,783 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.788e+02 3.447e+02 4.179e+02 1.001e+03, threshold=6.895e+02, percent-clipped=4.0 +2023-02-06 10:22:48,584 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84459.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:22:50,731 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84462.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:23:03,532 INFO [train.py:901] (1/4) Epoch 11, batch 3650, loss[loss=0.2601, simple_loss=0.3092, pruned_loss=0.1055, over 7974.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3123, pruned_loss=0.08208, over 1607250.54 frames. ], batch size: 21, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:23:11,597 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84493.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:23:12,298 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5329, 2.1578, 4.4725, 1.2748, 3.1004, 2.2803, 1.5853, 2.7084], + device='cuda:1'), covar=tensor([0.1716, 0.2247, 0.0577, 0.3847, 0.1435, 0.2606, 0.1859, 0.2333], + device='cuda:1'), in_proj_covar=tensor([0.0485, 0.0514, 0.0530, 0.0575, 0.0613, 0.0548, 0.0471, 0.0612], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 10:23:28,891 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84518.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:23:37,213 INFO [train.py:901] (1/4) Epoch 11, batch 3700, loss[loss=0.2215, simple_loss=0.2849, pruned_loss=0.07904, over 7659.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3126, pruned_loss=0.08221, over 1609726.18 frames. ], batch size: 19, lr: 6.98e-03, grad_scale: 8.0 +2023-02-06 10:23:46,126 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:23:48,582 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 10:23:49,860 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.734e+02 2.648e+02 3.219e+02 3.938e+02 7.332e+02, threshold=6.437e+02, percent-clipped=1.0 +2023-02-06 10:23:58,479 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84561.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:24:07,285 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84574.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:24:11,829 INFO [train.py:901] (1/4) Epoch 11, batch 3750, loss[loss=0.2619, simple_loss=0.3406, pruned_loss=0.09162, over 8658.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.313, pruned_loss=0.08175, over 1616105.25 frames. ], batch size: 39, lr: 6.98e-03, grad_scale: 8.0 +2023-02-06 10:24:23,993 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84598.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:24:46,862 INFO [train.py:901] (1/4) Epoch 11, batch 3800, loss[loss=0.3195, simple_loss=0.375, pruned_loss=0.132, over 8724.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3138, pruned_loss=0.08227, over 1612652.66 frames. ], batch size: 39, lr: 6.98e-03, grad_scale: 8.0 +2023-02-06 10:24:53,230 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.87 vs. limit=5.0 +2023-02-06 10:24:58,768 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.607e+02 3.118e+02 4.251e+02 1.041e+03, threshold=6.237e+02, percent-clipped=4.0 +2023-02-06 10:25:00,926 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84651.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:18,144 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84676.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:21,363 INFO [train.py:901] (1/4) Epoch 11, batch 3850, loss[loss=0.2574, simple_loss=0.3302, pruned_loss=0.09232, over 7057.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3134, pruned_loss=0.08218, over 1608853.39 frames. ], batch size: 72, lr: 6.98e-03, grad_scale: 8.0 +2023-02-06 10:25:22,265 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84682.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:25,124 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 +2023-02-06 10:25:25,853 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 10:25:32,819 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:39,748 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84707.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:43,784 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84713.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:47,047 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:51,556 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 10:25:55,495 INFO [train.py:901] (1/4) Epoch 11, batch 3900, loss[loss=0.2602, simple_loss=0.3367, pruned_loss=0.09185, over 8030.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.313, pruned_loss=0.08198, over 1607366.22 frames. ], batch size: 22, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:26:03,709 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84743.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:26:08,301 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.619e+02 3.238e+02 3.926e+02 9.069e+02, threshold=6.476e+02, percent-clipped=5.0 +2023-02-06 10:26:19,960 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5365, 1.8670, 2.0285, 1.1145, 2.0546, 1.2608, 0.5882, 1.6930], + device='cuda:1'), covar=tensor([0.0531, 0.0301, 0.0202, 0.0440, 0.0313, 0.0737, 0.0689, 0.0251], + device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0322, 0.0267, 0.0380, 0.0307, 0.0470, 0.0355, 0.0350], + device='cuda:1'), out_proj_covar=tensor([1.1065e-04, 9.0333e-05, 7.4837e-05, 1.0719e-04, 8.7397e-05, 1.4397e-04, + 1.0181e-04, 9.9744e-05], device='cuda:1') +2023-02-06 10:26:30,326 INFO [train.py:901] (1/4) Epoch 11, batch 3950, loss[loss=0.2149, simple_loss=0.285, pruned_loss=0.0724, over 7255.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3136, pruned_loss=0.08231, over 1609955.03 frames. ], batch size: 16, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:26:52,780 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:26:56,202 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84818.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:27:04,809 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84830.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:27:05,277 INFO [train.py:901] (1/4) Epoch 11, batch 4000, loss[loss=0.2218, simple_loss=0.2908, pruned_loss=0.07636, over 7522.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3135, pruned_loss=0.08196, over 1613408.39 frames. ], batch size: 18, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:27:07,061 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-06 10:27:17,173 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.608e+02 2.990e+02 3.694e+02 8.393e+02, threshold=5.981e+02, percent-clipped=2.0 +2023-02-06 10:27:21,544 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:27:39,667 INFO [train.py:901] (1/4) Epoch 11, batch 4050, loss[loss=0.1911, simple_loss=0.2698, pruned_loss=0.05621, over 7697.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3126, pruned_loss=0.08132, over 1610915.29 frames. ], batch size: 18, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:27:43,790 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84887.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:15,559 INFO [train.py:901] (1/4) Epoch 11, batch 4100, loss[loss=0.2641, simple_loss=0.3331, pruned_loss=0.09754, over 6933.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3123, pruned_loss=0.08145, over 1610032.25 frames. ], batch size: 72, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:28:16,466 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84932.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:26,691 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 +2023-02-06 10:28:27,736 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.518e+02 2.978e+02 3.788e+02 7.594e+02, threshold=5.956e+02, percent-clipped=4.0 +2023-02-06 10:28:33,385 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84957.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:34,132 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4330, 1.7278, 2.7340, 1.2543, 1.9179, 1.9517, 1.5304, 1.8646], + device='cuda:1'), covar=tensor([0.1788, 0.2155, 0.0825, 0.3690, 0.1615, 0.2673, 0.1889, 0.2079], + device='cuda:1'), in_proj_covar=tensor([0.0493, 0.0520, 0.0536, 0.0583, 0.0619, 0.0558, 0.0477, 0.0619], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 10:28:41,347 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:41,425 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:43,325 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0283, 2.3571, 1.9435, 2.9564, 1.3405, 1.7106, 1.8253, 2.3998], + device='cuda:1'), covar=tensor([0.0878, 0.0975, 0.1030, 0.0432, 0.1248, 0.1540, 0.1202, 0.0874], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0216, 0.0256, 0.0220, 0.0217, 0.0253, 0.0255, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 10:28:49,615 INFO [train.py:901] (1/4) Epoch 11, batch 4150, loss[loss=0.2215, simple_loss=0.278, pruned_loss=0.08256, over 7406.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3114, pruned_loss=0.08074, over 1613726.10 frames. ], batch size: 17, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:28:58,527 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84994.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:59,076 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84995.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:29:04,350 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85002.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:29:23,817 INFO [train.py:901] (1/4) Epoch 11, batch 4200, loss[loss=0.2738, simple_loss=0.3378, pruned_loss=0.1049, over 8466.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3118, pruned_loss=0.08098, over 1612690.95 frames. ], batch size: 25, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:29:36,444 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.682e+02 2.581e+02 3.261e+02 3.967e+02 9.417e+02, threshold=6.523e+02, percent-clipped=7.0 +2023-02-06 10:29:47,919 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 10:29:50,141 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85069.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:29:58,029 INFO [train.py:901] (1/4) Epoch 11, batch 4250, loss[loss=0.2767, simple_loss=0.3548, pruned_loss=0.09928, over 8358.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3108, pruned_loss=0.08055, over 1611980.42 frames. ], batch size: 24, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:30:06,842 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:30:10,060 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 10:30:18,069 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85110.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:30:25,268 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8858, 3.7840, 3.5161, 1.7365, 3.4228, 3.3684, 3.5664, 3.1098], + device='cuda:1'), covar=tensor([0.0827, 0.0609, 0.0927, 0.4590, 0.0807, 0.1056, 0.1210, 0.1062], + device='cuda:1'), in_proj_covar=tensor([0.0459, 0.0365, 0.0370, 0.0473, 0.0369, 0.0368, 0.0367, 0.0323], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 10:30:28,613 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:30:32,505 INFO [train.py:901] (1/4) Epoch 11, batch 4300, loss[loss=0.241, simple_loss=0.306, pruned_loss=0.08802, over 7807.00 frames. ], tot_loss[loss=0.236, simple_loss=0.3107, pruned_loss=0.08067, over 1613427.05 frames. ], batch size: 19, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:30:33,963 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85133.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:30:45,182 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.616e+02 3.014e+02 4.154e+02 7.931e+02, threshold=6.027e+02, percent-clipped=5.0 +2023-02-06 10:30:54,030 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85162.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:31:06,971 INFO [train.py:901] (1/4) Epoch 11, batch 4350, loss[loss=0.244, simple_loss=0.306, pruned_loss=0.09103, over 7782.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3118, pruned_loss=0.08156, over 1614619.45 frames. ], batch size: 19, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:31:36,560 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4925, 2.1665, 4.1254, 1.3509, 2.7880, 1.9954, 1.7086, 2.5419], + device='cuda:1'), covar=tensor([0.1924, 0.2486, 0.0722, 0.4235, 0.1774, 0.3259, 0.2020, 0.2761], + device='cuda:1'), in_proj_covar=tensor([0.0493, 0.0520, 0.0538, 0.0582, 0.0621, 0.0562, 0.0476, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 10:31:40,296 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 10:31:41,573 INFO [train.py:901] (1/4) Epoch 11, batch 4400, loss[loss=0.2396, simple_loss=0.3107, pruned_loss=0.08427, over 7931.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3128, pruned_loss=0.08216, over 1616518.61 frames. ], batch size: 20, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:31:54,340 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.553e+02 3.172e+02 3.669e+02 6.483e+02, threshold=6.345e+02, percent-clipped=4.0 +2023-02-06 10:32:01,427 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85258.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:32:14,074 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85277.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:32:16,485 INFO [train.py:901] (1/4) Epoch 11, batch 4450, loss[loss=0.2372, simple_loss=0.3152, pruned_loss=0.07955, over 8292.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3115, pruned_loss=0.08079, over 1615800.78 frames. ], batch size: 23, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:32:18,770 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:32:22,679 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 10:32:38,766 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85313.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:32:50,699 INFO [train.py:901] (1/4) Epoch 11, batch 4500, loss[loss=0.2198, simple_loss=0.3045, pruned_loss=0.06761, over 8694.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3109, pruned_loss=0.08039, over 1615702.34 frames. ], batch size: 34, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:33:00,922 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9349, 1.6502, 1.6747, 1.5461, 1.1177, 1.5481, 1.8209, 1.6972], + device='cuda:1'), covar=tensor([0.0515, 0.1122, 0.1570, 0.1257, 0.0581, 0.1405, 0.0666, 0.0556], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0154, 0.0193, 0.0159, 0.0105, 0.0164, 0.0118, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 10:33:03,403 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.629e+02 3.227e+02 4.085e+02 1.162e+03, threshold=6.455e+02, percent-clipped=2.0 +2023-02-06 10:33:15,858 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 10:33:16,083 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:33:26,545 INFO [train.py:901] (1/4) Epoch 11, batch 4550, loss[loss=0.1861, simple_loss=0.2857, pruned_loss=0.04324, over 8248.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.311, pruned_loss=0.08068, over 1618658.00 frames. ], batch size: 24, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:33:27,819 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 10:33:33,530 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85391.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:33:59,385 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:34:01,131 INFO [train.py:901] (1/4) Epoch 11, batch 4600, loss[loss=0.2409, simple_loss=0.3182, pruned_loss=0.08175, over 8455.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3105, pruned_loss=0.08056, over 1617403.07 frames. ], batch size: 25, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:34:11,928 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85446.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:34:13,797 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 2.573e+02 3.214e+02 4.149e+02 1.527e+03, threshold=6.427e+02, percent-clipped=2.0 +2023-02-06 10:34:14,010 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1434, 2.2989, 1.9197, 2.7340, 1.4332, 1.6912, 1.9806, 2.3493], + device='cuda:1'), covar=tensor([0.0703, 0.0765, 0.0971, 0.0394, 0.1105, 0.1304, 0.0886, 0.0753], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0217, 0.0259, 0.0219, 0.0220, 0.0257, 0.0259, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 10:34:27,393 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85469.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:34:33,579 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:34:36,247 INFO [train.py:901] (1/4) Epoch 11, batch 4650, loss[loss=0.2336, simple_loss=0.3055, pruned_loss=0.08086, over 7425.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3116, pruned_loss=0.08129, over 1615468.22 frames. ], batch size: 17, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:34:49,881 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.74 vs. limit=5.0 +2023-02-06 10:34:50,387 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85501.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:34:51,873 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9461, 2.2695, 3.6895, 1.7313, 2.9840, 2.3608, 2.0955, 2.7611], + device='cuda:1'), covar=tensor([0.1342, 0.1883, 0.0604, 0.3010, 0.1214, 0.2030, 0.1486, 0.1848], + device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0517, 0.0531, 0.0576, 0.0618, 0.0554, 0.0474, 0.0612], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 10:35:11,112 INFO [train.py:901] (1/4) Epoch 11, batch 4700, loss[loss=0.2835, simple_loss=0.3497, pruned_loss=0.1087, over 8518.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3125, pruned_loss=0.08204, over 1612414.01 frames. ], batch size: 26, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:35:12,715 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:35:22,561 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85548.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:35:23,054 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.812e+02 3.491e+02 4.674e+02 1.006e+03, threshold=6.983e+02, percent-clipped=9.0 +2023-02-06 10:35:30,008 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85558.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:35:45,766 INFO [train.py:901] (1/4) Epoch 11, batch 4750, loss[loss=0.1946, simple_loss=0.2632, pruned_loss=0.06301, over 7787.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3126, pruned_loss=0.08191, over 1618449.49 frames. ], batch size: 19, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:35:47,966 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85584.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:35:52,336 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 10:35:53,283 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:36:09,834 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 10:36:11,808 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 10:36:20,655 INFO [train.py:901] (1/4) Epoch 11, batch 4800, loss[loss=0.2267, simple_loss=0.3152, pruned_loss=0.06908, over 8104.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3137, pruned_loss=0.08232, over 1621225.56 frames. ], batch size: 23, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:36:28,779 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85643.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:36:32,782 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 2.628e+02 3.255e+02 4.281e+02 8.051e+02, threshold=6.510e+02, percent-clipped=3.0 +2023-02-06 10:36:35,795 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5488, 2.1199, 3.1024, 2.5582, 2.7705, 2.2680, 1.9989, 2.0259], + device='cuda:1'), covar=tensor([0.2819, 0.3419, 0.0939, 0.1858, 0.1506, 0.1835, 0.1457, 0.2991], + device='cuda:1'), in_proj_covar=tensor([0.0877, 0.0863, 0.0729, 0.0827, 0.0932, 0.0792, 0.0700, 0.0766], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 10:36:40,524 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1446, 2.2035, 1.7647, 2.0757, 1.7363, 1.3765, 1.6418, 1.7578], + device='cuda:1'), covar=tensor([0.1265, 0.0355, 0.0994, 0.0416, 0.0580, 0.1356, 0.0897, 0.0777], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0235, 0.0316, 0.0295, 0.0303, 0.0322, 0.0340, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 10:36:42,175 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-06 10:36:55,312 INFO [train.py:901] (1/4) Epoch 11, batch 4850, loss[loss=0.257, simple_loss=0.3144, pruned_loss=0.09984, over 8083.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3117, pruned_loss=0.081, over 1617338.17 frames. ], batch size: 21, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:36:57,567 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85684.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:37:01,293 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 10:37:14,640 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:37:30,196 INFO [train.py:901] (1/4) Epoch 11, batch 4900, loss[loss=0.2404, simple_loss=0.3257, pruned_loss=0.0776, over 8359.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3118, pruned_loss=0.08083, over 1614946.08 frames. ], batch size: 24, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:37:42,889 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 2.544e+02 3.151e+02 4.004e+02 8.063e+02, threshold=6.301e+02, percent-clipped=5.0 +2023-02-06 10:38:04,653 INFO [train.py:901] (1/4) Epoch 11, batch 4950, loss[loss=0.2307, simple_loss=0.316, pruned_loss=0.07273, over 8031.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3119, pruned_loss=0.08073, over 1615061.36 frames. ], batch size: 22, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:38:11,377 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85790.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:38:17,561 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.3741, 5.3917, 4.8142, 2.2599, 4.8257, 5.0007, 5.0833, 4.5281], + device='cuda:1'), covar=tensor([0.0617, 0.0421, 0.0808, 0.4572, 0.0769, 0.0748, 0.0967, 0.0728], + device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0374, 0.0379, 0.0484, 0.0378, 0.0378, 0.0375, 0.0334], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 10:38:28,576 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6048, 1.9008, 1.9468, 1.0990, 2.0708, 1.4970, 0.3516, 1.7113], + device='cuda:1'), covar=tensor([0.0319, 0.0184, 0.0159, 0.0315, 0.0208, 0.0554, 0.0546, 0.0148], + device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0322, 0.0269, 0.0380, 0.0308, 0.0470, 0.0355, 0.0347], + device='cuda:1'), out_proj_covar=tensor([1.1049e-04, 8.9705e-05, 7.5298e-05, 1.0694e-04, 8.7683e-05, 1.4383e-04, + 1.0160e-04, 9.8904e-05], device='cuda:1') +2023-02-06 10:38:39,599 INFO [train.py:901] (1/4) Epoch 11, batch 5000, loss[loss=0.1961, simple_loss=0.2821, pruned_loss=0.05503, over 8468.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3105, pruned_loss=0.08009, over 1614928.17 frames. ], batch size: 25, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:38:46,500 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85840.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:38:49,801 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85845.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:38:51,926 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:38:52,299 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.585e+02 3.219e+02 4.097e+02 8.363e+02, threshold=6.438e+02, percent-clipped=6.0 +2023-02-06 10:39:03,557 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85865.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:08,905 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85873.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:11,146 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-06 10:39:13,948 INFO [train.py:901] (1/4) Epoch 11, batch 5050, loss[loss=0.2222, simple_loss=0.2899, pruned_loss=0.07726, over 7639.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3112, pruned_loss=0.08054, over 1612697.65 frames. ], batch size: 19, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:39:21,267 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85892.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:22,011 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85893.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:30,139 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85905.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:39,818 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 10:39:48,589 INFO [train.py:901] (1/4) Epoch 11, batch 5100, loss[loss=0.2044, simple_loss=0.2789, pruned_loss=0.06495, over 7650.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3102, pruned_loss=0.07964, over 1612814.93 frames. ], batch size: 19, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:39:48,846 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8612, 1.9082, 2.2827, 1.6561, 1.1146, 2.5041, 0.2061, 1.3095], + device='cuda:1'), covar=tensor([0.2497, 0.1602, 0.0555, 0.2084, 0.4674, 0.0499, 0.3863, 0.2496], + device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0168, 0.0100, 0.0215, 0.0256, 0.0104, 0.0167, 0.0166], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 10:40:00,812 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85948.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:40:01,252 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.570e+02 3.113e+02 3.980e+02 6.838e+02, threshold=6.226e+02, percent-clipped=2.0 +2023-02-06 10:40:08,944 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85960.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:19,544 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85975.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:23,496 INFO [train.py:901] (1/4) Epoch 11, batch 5150, loss[loss=0.2695, simple_loss=0.3415, pruned_loss=0.09878, over 8283.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3121, pruned_loss=0.08052, over 1611355.44 frames. ], batch size: 23, lr: 6.92e-03, grad_scale: 8.0 +2023-02-06 10:40:26,411 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7701, 1.3809, 1.5157, 1.2723, 0.9804, 1.2408, 1.5970, 1.3738], + device='cuda:1'), covar=tensor([0.0525, 0.1279, 0.1752, 0.1453, 0.0578, 0.1554, 0.0688, 0.0619], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0151, 0.0191, 0.0158, 0.0103, 0.0163, 0.0116, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 10:40:27,614 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85987.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:42,275 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86007.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:51,391 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86020.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:59,434 INFO [train.py:901] (1/4) Epoch 11, batch 5200, loss[loss=0.2568, simple_loss=0.337, pruned_loss=0.08827, over 8495.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3113, pruned_loss=0.07975, over 1614434.14 frames. ], batch size: 26, lr: 6.92e-03, grad_scale: 16.0 +2023-02-06 10:41:12,339 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.648e+02 3.082e+02 3.913e+02 1.007e+03, threshold=6.165e+02, percent-clipped=5.0 +2023-02-06 10:41:27,743 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86070.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:41:35,330 INFO [train.py:901] (1/4) Epoch 11, batch 5250, loss[loss=0.2525, simple_loss=0.3328, pruned_loss=0.08607, over 8323.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3098, pruned_loss=0.07906, over 1614779.32 frames. ], batch size: 25, lr: 6.92e-03, grad_scale: 16.0 +2023-02-06 10:41:39,719 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:41:40,975 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 10:41:50,756 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86102.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:42:10,639 INFO [train.py:901] (1/4) Epoch 11, batch 5300, loss[loss=0.236, simple_loss=0.3123, pruned_loss=0.07986, over 8324.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3108, pruned_loss=0.07929, over 1618393.19 frames. ], batch size: 26, lr: 6.92e-03, grad_scale: 16.0 +2023-02-06 10:42:23,758 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.570e+02 3.118e+02 4.195e+02 8.045e+02, threshold=6.237e+02, percent-clipped=4.0 +2023-02-06 10:42:32,839 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:42:46,496 INFO [train.py:901] (1/4) Epoch 11, batch 5350, loss[loss=0.1909, simple_loss=0.2725, pruned_loss=0.05462, over 7797.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3105, pruned_loss=0.07971, over 1612791.94 frames. ], batch size: 19, lr: 6.92e-03, grad_scale: 16.0 +2023-02-06 10:42:50,703 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:12,424 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86216.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:22,274 INFO [train.py:901] (1/4) Epoch 11, batch 5400, loss[loss=0.2752, simple_loss=0.3521, pruned_loss=0.09919, over 8029.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3116, pruned_loss=0.0807, over 1613061.48 frames. ], batch size: 22, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:43:26,576 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86237.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:29,462 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86241.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:34,672 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.471e+02 3.223e+02 4.268e+02 9.619e+02, threshold=6.446e+02, percent-clipped=7.0 +2023-02-06 10:43:44,506 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86263.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:45,131 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86264.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:57,396 INFO [train.py:901] (1/4) Epoch 11, batch 5450, loss[loss=0.2733, simple_loss=0.3391, pruned_loss=0.1037, over 8504.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3117, pruned_loss=0.08081, over 1617304.12 frames. ], batch size: 28, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:44:03,077 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:44:05,824 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86292.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:44:23,287 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.57 vs. limit=5.0 +2023-02-06 10:44:25,144 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86319.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:44:34,087 INFO [train.py:901] (1/4) Epoch 11, batch 5500, loss[loss=0.2441, simple_loss=0.3257, pruned_loss=0.0812, over 8333.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3109, pruned_loss=0.08026, over 1613510.30 frames. ], batch size: 25, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:44:34,728 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 10:44:37,215 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-06 10:44:46,146 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.494e+02 3.013e+02 3.770e+02 8.759e+02, threshold=6.025e+02, percent-clipped=3.0 +2023-02-06 10:44:48,475 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86352.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:44:52,710 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:44:56,787 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:44:58,664 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-06 10:45:09,177 INFO [train.py:901] (1/4) Epoch 11, batch 5550, loss[loss=0.2283, simple_loss=0.2967, pruned_loss=0.07993, over 7932.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3116, pruned_loss=0.08039, over 1614173.14 frames. ], batch size: 20, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:45:10,784 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86383.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:45:27,857 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86407.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:45:33,158 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86414.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:45:44,323 INFO [train.py:901] (1/4) Epoch 11, batch 5600, loss[loss=0.2455, simple_loss=0.321, pruned_loss=0.08505, over 8758.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3113, pruned_loss=0.0803, over 1614992.90 frames. ], batch size: 30, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:45:44,401 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86431.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:45:46,511 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:45:51,968 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0195, 1.4717, 3.2498, 1.2045, 2.3104, 3.5977, 3.6052, 3.0422], + device='cuda:1'), covar=tensor([0.1121, 0.1688, 0.0371, 0.2282, 0.0942, 0.0233, 0.0492, 0.0618], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0296, 0.0257, 0.0288, 0.0269, 0.0235, 0.0337, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 10:45:57,241 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.527e+02 3.003e+02 3.802e+02 9.548e+02, threshold=6.005e+02, percent-clipped=4.0 +2023-02-06 10:46:03,247 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86458.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:46:06,580 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86463.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:46:17,335 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:46:18,539 INFO [train.py:901] (1/4) Epoch 11, batch 5650, loss[loss=0.1927, simple_loss=0.2813, pruned_loss=0.05207, over 8194.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3113, pruned_loss=0.08071, over 1612355.67 frames. ], batch size: 23, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:46:39,890 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 10:46:48,926 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3921, 1.9727, 2.9242, 2.3003, 2.5224, 2.2347, 1.8370, 1.2606], + device='cuda:1'), covar=tensor([0.3888, 0.4004, 0.1147, 0.2466, 0.1976, 0.2152, 0.1709, 0.4233], + device='cuda:1'), in_proj_covar=tensor([0.0885, 0.0862, 0.0724, 0.0831, 0.0929, 0.0792, 0.0698, 0.0762], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 10:46:52,197 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86529.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:46:53,980 INFO [train.py:901] (1/4) Epoch 11, batch 5700, loss[loss=0.2776, simple_loss=0.3432, pruned_loss=0.106, over 8521.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3125, pruned_loss=0.08169, over 1614497.60 frames. ], batch size: 26, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:47:04,209 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86546.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:47:06,038 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.473e+02 3.032e+02 3.837e+02 8.433e+02, threshold=6.065e+02, percent-clipped=5.0 +2023-02-06 10:47:28,602 INFO [train.py:901] (1/4) Epoch 11, batch 5750, loss[loss=0.2117, simple_loss=0.2772, pruned_loss=0.07313, over 7817.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3114, pruned_loss=0.08082, over 1613366.55 frames. ], batch size: 20, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:47:40,221 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86598.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:47:42,139 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 10:47:47,614 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:47:47,759 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:47:54,772 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 10:48:00,607 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2145, 1.7509, 2.5698, 1.9864, 2.2795, 2.0484, 1.7036, 0.9410], + device='cuda:1'), covar=tensor([0.3815, 0.3661, 0.1095, 0.2464, 0.1749, 0.2238, 0.1680, 0.3853], + device='cuda:1'), in_proj_covar=tensor([0.0884, 0.0866, 0.0727, 0.0835, 0.0930, 0.0794, 0.0699, 0.0764], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 10:48:03,810 INFO [train.py:901] (1/4) Epoch 11, batch 5800, loss[loss=0.2233, simple_loss=0.2872, pruned_loss=0.07972, over 7216.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3095, pruned_loss=0.08024, over 1606702.50 frames. ], batch size: 16, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:48:05,399 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86633.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:48:13,803 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3284, 2.9025, 2.3981, 3.7513, 1.6977, 2.0558, 2.2233, 2.9701], + device='cuda:1'), covar=tensor([0.0794, 0.0731, 0.0890, 0.0300, 0.1162, 0.1384, 0.1201, 0.0798], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0214, 0.0252, 0.0217, 0.0218, 0.0254, 0.0254, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 10:48:17,062 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 2.625e+02 3.434e+02 4.363e+02 1.044e+03, threshold=6.867e+02, percent-clipped=16.0 +2023-02-06 10:48:26,845 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86663.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:48:39,359 INFO [train.py:901] (1/4) Epoch 11, batch 5850, loss[loss=0.2578, simple_loss=0.3361, pruned_loss=0.0897, over 8246.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3094, pruned_loss=0.08026, over 1602168.40 frames. ], batch size: 24, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:48:44,282 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86688.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:48:45,640 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86690.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:02,117 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:07,940 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86723.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:08,185 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 10:49:13,215 INFO [train.py:901] (1/4) Epoch 11, batch 5900, loss[loss=0.2189, simple_loss=0.2826, pruned_loss=0.07761, over 7203.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3104, pruned_loss=0.08096, over 1603445.37 frames. ], batch size: 16, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:49:16,680 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86735.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:25,724 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.878e+02 2.650e+02 3.002e+02 3.837e+02 8.505e+02, threshold=6.004e+02, percent-clipped=1.0 +2023-02-06 10:49:33,355 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:48,259 INFO [train.py:901] (1/4) Epoch 11, batch 5950, loss[loss=0.2213, simple_loss=0.2934, pruned_loss=0.07459, over 7982.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3102, pruned_loss=0.08066, over 1606070.58 frames. ], batch size: 21, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:49:51,984 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:57,466 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4608, 2.0016, 3.1620, 2.4142, 2.7546, 2.2856, 1.7954, 1.4073], + device='cuda:1'), covar=tensor([0.3987, 0.4010, 0.1092, 0.2565, 0.1872, 0.2132, 0.1634, 0.4281], + device='cuda:1'), in_proj_covar=tensor([0.0887, 0.0869, 0.0728, 0.0839, 0.0930, 0.0798, 0.0700, 0.0765], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 10:50:03,635 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:50:03,784 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:50:06,870 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:50:09,065 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:50:10,679 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 10:50:20,474 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86827.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:50:22,932 INFO [train.py:901] (1/4) Epoch 11, batch 6000, loss[loss=0.2624, simple_loss=0.3359, pruned_loss=0.09442, over 8477.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3093, pruned_loss=0.0802, over 1604617.49 frames. ], batch size: 25, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:50:22,932 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 10:50:35,331 INFO [train.py:935] (1/4) Epoch 11, validation: loss=0.1887, simple_loss=0.2887, pruned_loss=0.04439, over 944034.00 frames. +2023-02-06 10:50:35,332 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 10:50:36,204 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86832.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:50:47,363 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.431e+02 2.934e+02 3.566e+02 7.044e+02, threshold=5.869e+02, percent-clipped=5.0 +2023-02-06 10:51:10,323 INFO [train.py:901] (1/4) Epoch 11, batch 6050, loss[loss=0.2137, simple_loss=0.2781, pruned_loss=0.07463, over 8028.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3104, pruned_loss=0.08116, over 1608556.97 frames. ], batch size: 22, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:51:20,528 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86896.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:51:35,592 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86917.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:51:39,111 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86922.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:51:45,230 INFO [train.py:901] (1/4) Epoch 11, batch 6100, loss[loss=0.2261, simple_loss=0.3054, pruned_loss=0.07339, over 8537.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.31, pruned_loss=0.08062, over 1610942.72 frames. ], batch size: 50, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:51:53,492 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86942.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:51:58,241 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.604e+02 3.114e+02 3.901e+02 9.212e+02, threshold=6.229e+02, percent-clipped=4.0 +2023-02-06 10:52:06,861 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 10:52:19,160 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:52:20,335 INFO [train.py:901] (1/4) Epoch 11, batch 6150, loss[loss=0.2722, simple_loss=0.3444, pruned_loss=0.09996, over 8501.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3094, pruned_loss=0.07962, over 1612398.98 frames. ], batch size: 26, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:52:36,919 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87004.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:52:47,743 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87020.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:52:55,885 INFO [train.py:901] (1/4) Epoch 11, batch 6200, loss[loss=0.1857, simple_loss=0.2711, pruned_loss=0.05013, over 8462.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.309, pruned_loss=0.07935, over 1610676.94 frames. ], batch size: 25, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:53:06,093 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8283, 2.0950, 1.6522, 2.6801, 1.3186, 1.4723, 1.7683, 2.2248], + device='cuda:1'), covar=tensor([0.0844, 0.0940, 0.1182, 0.0391, 0.1104, 0.1502, 0.0991, 0.0888], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0215, 0.0256, 0.0217, 0.0218, 0.0253, 0.0253, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 10:53:07,901 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.592e+02 3.192e+02 4.476e+02 1.804e+03, threshold=6.384e+02, percent-clipped=5.0 +2023-02-06 10:53:14,474 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:53:30,924 INFO [train.py:901] (1/4) Epoch 11, batch 6250, loss[loss=0.2231, simple_loss=0.309, pruned_loss=0.06865, over 5064.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3103, pruned_loss=0.07996, over 1611655.88 frames. ], batch size: 11, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:54:06,553 INFO [train.py:901] (1/4) Epoch 11, batch 6300, loss[loss=0.2278, simple_loss=0.3066, pruned_loss=0.07449, over 8595.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3094, pruned_loss=0.0798, over 1607274.50 frames. ], batch size: 49, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:54:19,292 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.563e+02 3.017e+02 3.734e+02 8.364e+02, threshold=6.034e+02, percent-clipped=3.0 +2023-02-06 10:54:36,433 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87173.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:54:38,306 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87176.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:54:39,802 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87178.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:54:41,629 INFO [train.py:901] (1/4) Epoch 11, batch 6350, loss[loss=0.2336, simple_loss=0.2891, pruned_loss=0.08908, over 7705.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3104, pruned_loss=0.08043, over 1609195.30 frames. ], batch size: 18, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:54:44,729 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 10:54:53,189 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87198.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:54:57,275 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87203.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:55:16,801 INFO [train.py:901] (1/4) Epoch 11, batch 6400, loss[loss=0.1979, simple_loss=0.2722, pruned_loss=0.06178, over 7707.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3102, pruned_loss=0.08062, over 1610252.94 frames. ], batch size: 18, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:55:23,183 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87240.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:55:28,866 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87248.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:55:29,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.810e+02 2.577e+02 3.020e+02 3.786e+02 7.428e+02, threshold=6.041e+02, percent-clipped=2.0 +2023-02-06 10:55:37,016 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3888, 2.0602, 3.3132, 1.2178, 2.6349, 1.8588, 1.5709, 2.3845], + device='cuda:1'), covar=tensor([0.1858, 0.2257, 0.0873, 0.4086, 0.1541, 0.2924, 0.2077, 0.2171], + device='cuda:1'), in_proj_covar=tensor([0.0486, 0.0523, 0.0541, 0.0580, 0.0619, 0.0557, 0.0473, 0.0616], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 10:55:51,527 INFO [train.py:901] (1/4) Epoch 11, batch 6450, loss[loss=0.2443, simple_loss=0.319, pruned_loss=0.08477, over 8025.00 frames. ], tot_loss[loss=0.236, simple_loss=0.3104, pruned_loss=0.0808, over 1610690.86 frames. ], batch size: 22, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:55:59,185 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:56:14,144 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87313.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:56:14,840 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3175, 2.4849, 1.6726, 2.0106, 1.8732, 1.3067, 1.7837, 1.8722], + device='cuda:1'), covar=tensor([0.1330, 0.0325, 0.1092, 0.0627, 0.0655, 0.1391, 0.0932, 0.0858], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0235, 0.0316, 0.0296, 0.0302, 0.0320, 0.0340, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 10:56:17,607 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3004, 1.7997, 2.7615, 2.1751, 2.4645, 2.1040, 1.7484, 1.2855], + device='cuda:1'), covar=tensor([0.3757, 0.4094, 0.1170, 0.2504, 0.1799, 0.2244, 0.1648, 0.4030], + device='cuda:1'), in_proj_covar=tensor([0.0884, 0.0866, 0.0726, 0.0840, 0.0925, 0.0794, 0.0698, 0.0761], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 10:56:27,327 INFO [train.py:901] (1/4) Epoch 11, batch 6500, loss[loss=0.2617, simple_loss=0.3395, pruned_loss=0.09196, over 8497.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3106, pruned_loss=0.08063, over 1613012.13 frames. ], batch size: 26, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:56:32,359 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:56:39,861 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.605e+02 3.245e+02 4.169e+02 7.875e+02, threshold=6.489e+02, percent-clipped=5.0 +2023-02-06 10:56:44,232 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87355.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:56:50,430 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87364.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:57:01,903 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87380.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:57:02,442 INFO [train.py:901] (1/4) Epoch 11, batch 6550, loss[loss=0.2293, simple_loss=0.2964, pruned_loss=0.08107, over 7801.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3109, pruned_loss=0.08017, over 1619609.55 frames. ], batch size: 20, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:57:17,764 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 10:57:36,586 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1570, 2.4216, 1.9360, 3.0246, 1.5283, 1.6200, 1.9693, 2.4960], + device='cuda:1'), covar=tensor([0.0750, 0.0839, 0.0988, 0.0308, 0.1065, 0.1408, 0.0903, 0.0820], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0215, 0.0258, 0.0218, 0.0220, 0.0257, 0.0258, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 10:57:37,077 INFO [train.py:901] (1/4) Epoch 11, batch 6600, loss[loss=0.1878, simple_loss=0.2493, pruned_loss=0.06316, over 7524.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3109, pruned_loss=0.08015, over 1618084.36 frames. ], batch size: 18, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:57:37,783 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 10:57:41,254 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4218, 4.3112, 4.0149, 1.7537, 3.8420, 3.9817, 3.9734, 3.7320], + device='cuda:1'), covar=tensor([0.0774, 0.0606, 0.0967, 0.5259, 0.0792, 0.0985, 0.1262, 0.0772], + device='cuda:1'), in_proj_covar=tensor([0.0461, 0.0372, 0.0382, 0.0481, 0.0379, 0.0375, 0.0377, 0.0333], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 10:57:46,876 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9055, 1.9046, 2.2950, 1.7256, 1.2833, 2.5722, 0.6110, 1.4939], + device='cuda:1'), covar=tensor([0.2379, 0.1912, 0.0554, 0.2094, 0.4275, 0.0384, 0.3412, 0.2005], + device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0167, 0.0099, 0.0213, 0.0250, 0.0101, 0.0160, 0.0163], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 10:57:47,567 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4787, 1.8343, 1.7515, 1.1213, 1.8940, 1.3345, 0.5548, 1.6819], + device='cuda:1'), covar=tensor([0.0330, 0.0192, 0.0159, 0.0321, 0.0221, 0.0550, 0.0483, 0.0163], + device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0323, 0.0266, 0.0381, 0.0308, 0.0466, 0.0349, 0.0344], + device='cuda:1'), out_proj_covar=tensor([1.1076e-04, 9.0009e-05, 7.4450e-05, 1.0723e-04, 8.7297e-05, 1.4185e-04, + 9.9638e-05, 9.7645e-05], device='cuda:1') +2023-02-06 10:57:50,083 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.293e+02 2.790e+02 3.732e+02 8.562e+02, threshold=5.581e+02, percent-clipped=1.0 +2023-02-06 10:57:55,854 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-06 10:58:11,498 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:58:11,537 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87479.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:58:12,692 INFO [train.py:901] (1/4) Epoch 11, batch 6650, loss[loss=0.205, simple_loss=0.2801, pruned_loss=0.06496, over 7692.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3104, pruned_loss=0.0798, over 1619873.40 frames. ], batch size: 18, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 10:58:18,142 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8894, 2.0489, 1.6772, 2.5150, 1.2427, 1.5597, 1.6819, 2.0061], + device='cuda:1'), covar=tensor([0.0715, 0.0797, 0.1040, 0.0413, 0.1186, 0.1306, 0.0933, 0.0829], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0214, 0.0257, 0.0217, 0.0219, 0.0254, 0.0257, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 10:58:41,497 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87523.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:58:47,482 INFO [train.py:901] (1/4) Epoch 11, batch 6700, loss[loss=0.2402, simple_loss=0.298, pruned_loss=0.09114, over 7795.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3098, pruned_loss=0.07949, over 1619345.74 frames. ], batch size: 19, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 10:58:58,342 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87547.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:58:59,459 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.493e+02 3.158e+02 4.170e+02 8.693e+02, threshold=6.316e+02, percent-clipped=8.0 +2023-02-06 10:59:16,931 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87572.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:22,976 INFO [train.py:901] (1/4) Epoch 11, batch 6750, loss[loss=0.1991, simple_loss=0.2867, pruned_loss=0.05579, over 8234.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3098, pruned_loss=0.07994, over 1618314.93 frames. ], batch size: 22, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 10:59:30,584 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:37,575 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87602.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:39,609 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4570, 1.6335, 2.4062, 1.3120, 1.6738, 1.7643, 1.4334, 1.5455], + device='cuda:1'), covar=tensor([0.1595, 0.1924, 0.0600, 0.3592, 0.1315, 0.2625, 0.1869, 0.1559], + device='cuda:1'), in_proj_covar=tensor([0.0479, 0.0515, 0.0525, 0.0570, 0.0608, 0.0544, 0.0464, 0.0602], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 10:59:40,789 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:43,430 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87611.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:52,275 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:56,919 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 10:59:57,597 INFO [train.py:901] (1/4) Epoch 11, batch 6800, loss[loss=0.2974, simple_loss=0.3522, pruned_loss=0.1213, over 6734.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3097, pruned_loss=0.08012, over 1616514.13 frames. ], batch size: 72, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 11:00:01,245 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:00:01,931 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9193, 2.7290, 3.4598, 1.7772, 1.5539, 3.7168, 0.4874, 1.9510], + device='cuda:1'), covar=tensor([0.2104, 0.1618, 0.0455, 0.3511, 0.4633, 0.0270, 0.3830, 0.2348], + device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0163, 0.0097, 0.0209, 0.0244, 0.0100, 0.0157, 0.0160], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 11:00:10,522 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 2.375e+02 2.980e+02 3.798e+02 7.616e+02, threshold=5.961e+02, percent-clipped=2.0 +2023-02-06 11:00:23,987 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8437, 2.6900, 3.1842, 2.1868, 1.6749, 3.4013, 0.5317, 2.1155], + device='cuda:1'), covar=tensor([0.2160, 0.1263, 0.0438, 0.2346, 0.3932, 0.0429, 0.3558, 0.1867], + device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0165, 0.0098, 0.0212, 0.0248, 0.0101, 0.0159, 0.0162], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 11:00:32,370 INFO [train.py:901] (1/4) Epoch 11, batch 6850, loss[loss=0.2021, simple_loss=0.2818, pruned_loss=0.06123, over 7933.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3098, pruned_loss=0.08046, over 1618578.91 frames. ], batch size: 20, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 11:00:45,129 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 11:00:50,747 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87707.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:00:51,408 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6697, 1.3858, 4.8064, 1.6752, 4.2288, 3.9794, 4.3518, 4.1673], + device='cuda:1'), covar=tensor([0.0457, 0.4422, 0.0413, 0.3517, 0.0994, 0.0828, 0.0460, 0.0589], + device='cuda:1'), in_proj_covar=tensor([0.0486, 0.0571, 0.0579, 0.0529, 0.0596, 0.0514, 0.0505, 0.0572], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:01:01,872 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87724.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:01:06,419 INFO [train.py:901] (1/4) Epoch 11, batch 6900, loss[loss=0.2325, simple_loss=0.3069, pruned_loss=0.07903, over 8032.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3103, pruned_loss=0.08054, over 1616931.54 frames. ], batch size: 22, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 11:01:10,021 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87735.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:01:19,188 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.628e+02 3.043e+02 4.130e+02 7.700e+02, threshold=6.086e+02, percent-clipped=2.0 +2023-02-06 11:01:22,687 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7387, 3.6855, 3.3948, 1.7325, 3.3143, 3.2891, 3.4205, 3.0491], + device='cuda:1'), covar=tensor([0.1115, 0.0798, 0.1244, 0.4991, 0.1021, 0.1187, 0.1620, 0.1124], + device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0374, 0.0384, 0.0480, 0.0377, 0.0376, 0.0377, 0.0336], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:01:26,779 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8393, 1.4258, 5.9919, 2.1038, 5.3077, 4.9611, 5.5515, 5.3901], + device='cuda:1'), covar=tensor([0.0478, 0.4982, 0.0324, 0.3416, 0.0969, 0.0827, 0.0444, 0.0490], + device='cuda:1'), in_proj_covar=tensor([0.0490, 0.0576, 0.0584, 0.0533, 0.0602, 0.0517, 0.0507, 0.0577], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:01:26,848 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87760.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:01:41,613 INFO [train.py:901] (1/4) Epoch 11, batch 6950, loss[loss=0.1972, simple_loss=0.2673, pruned_loss=0.06355, over 7694.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3094, pruned_loss=0.07981, over 1611802.63 frames. ], batch size: 18, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:01:52,557 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 11:02:11,622 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87823.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:02:16,826 INFO [train.py:901] (1/4) Epoch 11, batch 7000, loss[loss=0.2222, simple_loss=0.2983, pruned_loss=0.07307, over 7975.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3094, pruned_loss=0.0798, over 1612728.57 frames. ], batch size: 21, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:02:22,315 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87839.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:02:29,505 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.548e+02 3.185e+02 4.052e+02 9.283e+02, threshold=6.369e+02, percent-clipped=6.0 +2023-02-06 11:02:41,530 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87867.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:02:51,622 INFO [train.py:901] (1/4) Epoch 11, batch 7050, loss[loss=0.2652, simple_loss=0.3243, pruned_loss=0.103, over 7915.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3099, pruned_loss=0.07984, over 1614412.62 frames. ], batch size: 20, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:03:12,208 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5721, 1.6443, 2.0138, 1.7055, 1.0540, 2.0754, 0.2250, 1.2916], + device='cuda:1'), covar=tensor([0.2590, 0.1818, 0.0433, 0.1457, 0.4334, 0.0424, 0.3225, 0.1674], + device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0165, 0.0098, 0.0211, 0.0247, 0.0100, 0.0159, 0.0162], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 11:03:26,708 INFO [train.py:901] (1/4) Epoch 11, batch 7100, loss[loss=0.2449, simple_loss=0.3177, pruned_loss=0.08611, over 8301.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3103, pruned_loss=0.08004, over 1612444.56 frames. ], batch size: 23, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:03:31,616 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87938.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:03:36,825 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87946.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:03:38,774 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.732e+02 3.356e+02 4.654e+02 1.650e+03, threshold=6.712e+02, percent-clipped=12.0 +2023-02-06 11:03:40,145 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:03:48,276 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87963.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:03:51,539 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87967.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:04:00,737 INFO [train.py:901] (1/4) Epoch 11, batch 7150, loss[loss=0.2524, simple_loss=0.323, pruned_loss=0.09093, over 7166.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.31, pruned_loss=0.07982, over 1612891.43 frames. ], batch size: 16, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:04:01,614 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87982.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:04:05,825 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87988.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:04:36,666 INFO [train.py:901] (1/4) Epoch 11, batch 7200, loss[loss=0.2397, simple_loss=0.3152, pruned_loss=0.08209, over 6864.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3103, pruned_loss=0.08029, over 1611128.19 frames. ], batch size: 72, lr: 6.84e-03, grad_scale: 32.0 +2023-02-06 11:04:49,438 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.591e+02 3.086e+02 3.706e+02 9.715e+02, threshold=6.172e+02, percent-clipped=2.0 +2023-02-06 11:04:57,787 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88061.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:01,268 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:11,817 INFO [train.py:901] (1/4) Epoch 11, batch 7250, loss[loss=0.18, simple_loss=0.2558, pruned_loss=0.05211, over 7188.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3114, pruned_loss=0.08118, over 1608840.23 frames. ], batch size: 16, lr: 6.84e-03, grad_scale: 32.0 +2023-02-06 11:05:12,647 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:21,491 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88095.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:37,918 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88118.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:39,343 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:40,258 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.92 vs. limit=5.0 +2023-02-06 11:05:46,910 INFO [train.py:901] (1/4) Epoch 11, batch 7300, loss[loss=0.2178, simple_loss=0.3084, pruned_loss=0.0636, over 8142.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3106, pruned_loss=0.08089, over 1608089.67 frames. ], batch size: 22, lr: 6.84e-03, grad_scale: 32.0 +2023-02-06 11:06:00,692 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.414e+02 2.958e+02 3.757e+02 7.369e+02, threshold=5.915e+02, percent-clipped=2.0 +2023-02-06 11:06:10,194 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 11:06:22,812 INFO [train.py:901] (1/4) Epoch 11, batch 7350, loss[loss=0.2321, simple_loss=0.3162, pruned_loss=0.07393, over 8361.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3102, pruned_loss=0.08024, over 1608253.61 frames. ], batch size: 24, lr: 6.84e-03, grad_scale: 32.0 +2023-02-06 11:06:32,263 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 11:06:32,470 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88194.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:06:50,354 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88219.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:06:51,484 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 11:06:58,209 INFO [train.py:901] (1/4) Epoch 11, batch 7400, loss[loss=0.2371, simple_loss=0.3144, pruned_loss=0.07989, over 7980.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.31, pruned_loss=0.07959, over 1611083.66 frames. ], batch size: 21, lr: 6.84e-03, grad_scale: 16.0 +2023-02-06 11:07:03,173 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88238.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:07:11,891 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.577e+02 3.074e+02 3.691e+02 9.024e+02, threshold=6.148e+02, percent-clipped=4.0 +2023-02-06 11:07:20,895 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88263.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:07:26,471 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 11:07:31,794 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6227, 2.0101, 3.4187, 1.3623, 2.5503, 2.0809, 1.7405, 2.2487], + device='cuda:1'), covar=tensor([0.1670, 0.2164, 0.0761, 0.3761, 0.1458, 0.2615, 0.1757, 0.2231], + device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0522, 0.0529, 0.0576, 0.0618, 0.0558, 0.0472, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:07:32,914 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 11:07:33,599 INFO [train.py:901] (1/4) Epoch 11, batch 7450, loss[loss=0.2532, simple_loss=0.3277, pruned_loss=0.08939, over 7300.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.31, pruned_loss=0.07968, over 1611598.19 frames. ], batch size: 72, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:07:52,585 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88309.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:07:58,568 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88317.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:01,846 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:07,739 INFO [train.py:901] (1/4) Epoch 11, batch 7500, loss[loss=0.1946, simple_loss=0.2797, pruned_loss=0.05477, over 8103.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3096, pruned_loss=0.07961, over 1612068.25 frames. ], batch size: 23, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:08:13,329 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:15,895 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:19,206 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:20,937 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 2.827e+02 3.509e+02 4.304e+02 1.282e+03, threshold=7.018e+02, percent-clipped=8.0 +2023-02-06 11:08:30,007 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88363.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:42,578 INFO [train.py:901] (1/4) Epoch 11, batch 7550, loss[loss=0.2072, simple_loss=0.2809, pruned_loss=0.06678, over 7654.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3093, pruned_loss=0.08, over 1610335.27 frames. ], batch size: 19, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:08:48,011 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9994, 1.6049, 1.6668, 1.6178, 1.0802, 1.7006, 2.1438, 1.9147], + device='cuda:1'), covar=tensor([0.0414, 0.1238, 0.1746, 0.1340, 0.0632, 0.1473, 0.0659, 0.0550], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0150, 0.0190, 0.0157, 0.0103, 0.0164, 0.0115, 0.0136], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 11:09:03,019 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 11:09:13,612 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5854, 1.4108, 2.7712, 1.2363, 1.9998, 3.0237, 3.1378, 2.5937], + device='cuda:1'), covar=tensor([0.1087, 0.1458, 0.0425, 0.2012, 0.0948, 0.0297, 0.0546, 0.0599], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0298, 0.0261, 0.0292, 0.0273, 0.0236, 0.0338, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 11:09:17,367 INFO [train.py:901] (1/4) Epoch 11, batch 7600, loss[loss=0.2638, simple_loss=0.3338, pruned_loss=0.09687, over 8137.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.31, pruned_loss=0.08056, over 1613067.15 frames. ], batch size: 22, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:09:31,023 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.301e+02 2.441e+02 2.975e+02 3.888e+02 6.138e+02, threshold=5.951e+02, percent-clipped=0.0 +2023-02-06 11:09:39,145 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88462.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:09:51,499 INFO [train.py:901] (1/4) Epoch 11, batch 7650, loss[loss=0.2683, simple_loss=0.3396, pruned_loss=0.09846, over 8486.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3107, pruned_loss=0.08038, over 1614809.40 frames. ], batch size: 28, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:10:12,791 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 11:10:26,446 INFO [train.py:901] (1/4) Epoch 11, batch 7700, loss[loss=0.1924, simple_loss=0.2787, pruned_loss=0.05301, over 8081.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3109, pruned_loss=0.0807, over 1613299.17 frames. ], batch size: 21, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:10:38,717 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7075, 2.1449, 2.2704, 1.3756, 2.3958, 1.5962, 0.7395, 1.8064], + device='cuda:1'), covar=tensor([0.0448, 0.0196, 0.0160, 0.0382, 0.0245, 0.0571, 0.0559, 0.0219], + device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0324, 0.0269, 0.0378, 0.0308, 0.0468, 0.0355, 0.0348], + device='cuda:1'), out_proj_covar=tensor([1.0997e-04, 9.0117e-05, 7.4909e-05, 1.0604e-04, 8.6932e-05, 1.4251e-04, + 1.0098e-04, 9.8654e-05], device='cuda:1') +2023-02-06 11:10:39,159 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 11:10:39,711 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.472e+02 3.053e+02 3.571e+02 8.603e+02, threshold=6.105e+02, percent-clipped=3.0 +2023-02-06 11:10:58,414 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88577.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:11:00,812 INFO [train.py:901] (1/4) Epoch 11, batch 7750, loss[loss=0.1908, simple_loss=0.276, pruned_loss=0.05281, over 7430.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.311, pruned_loss=0.08038, over 1613997.24 frames. ], batch size: 17, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:11:20,952 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2390, 1.7995, 2.6345, 2.0676, 2.3333, 2.0695, 1.7297, 1.0096], + device='cuda:1'), covar=tensor([0.3955, 0.3898, 0.1190, 0.2399, 0.1710, 0.2130, 0.1644, 0.3884], + device='cuda:1'), in_proj_covar=tensor([0.0884, 0.0864, 0.0730, 0.0842, 0.0927, 0.0800, 0.0701, 0.0763], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:11:36,342 INFO [train.py:901] (1/4) Epoch 11, batch 7800, loss[loss=0.2257, simple_loss=0.3107, pruned_loss=0.07037, over 8467.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3094, pruned_loss=0.07984, over 1613451.92 frames. ], batch size: 25, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:11:42,970 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7460, 1.4403, 5.8573, 2.0967, 5.1842, 4.9368, 5.4238, 5.2014], + device='cuda:1'), covar=tensor([0.0464, 0.4524, 0.0350, 0.3180, 0.0889, 0.0680, 0.0389, 0.0467], + device='cuda:1'), in_proj_covar=tensor([0.0490, 0.0576, 0.0586, 0.0528, 0.0603, 0.0514, 0.0507, 0.0577], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:11:48,834 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.685e+02 3.345e+02 4.152e+02 1.012e+03, threshold=6.690e+02, percent-clipped=6.0 +2023-02-06 11:11:50,872 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88653.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:12:09,483 INFO [train.py:901] (1/4) Epoch 11, batch 7850, loss[loss=0.266, simple_loss=0.3457, pruned_loss=0.09316, over 8263.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3111, pruned_loss=0.08093, over 1613916.43 frames. ], batch size: 24, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:12:11,209 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-02-06 11:12:42,909 INFO [train.py:901] (1/4) Epoch 11, batch 7900, loss[loss=0.2582, simple_loss=0.3289, pruned_loss=0.09375, over 8190.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3108, pruned_loss=0.08073, over 1612856.57 frames. ], batch size: 23, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:12:55,426 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.490e+02 3.060e+02 3.735e+02 6.734e+02, threshold=6.120e+02, percent-clipped=1.0 +2023-02-06 11:12:58,825 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8314, 1.8292, 2.3615, 1.7060, 1.1850, 2.5513, 0.4113, 1.5186], + device='cuda:1'), covar=tensor([0.2889, 0.2436, 0.0572, 0.2551, 0.4908, 0.0557, 0.4025, 0.2072], + device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0168, 0.0101, 0.0212, 0.0252, 0.0103, 0.0163, 0.0165], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 11:13:07,261 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88768.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:13:15,811 INFO [train.py:901] (1/4) Epoch 11, batch 7950, loss[loss=0.2019, simple_loss=0.2836, pruned_loss=0.06008, over 7791.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3119, pruned_loss=0.08121, over 1613239.22 frames. ], batch size: 19, lr: 6.81e-03, grad_scale: 16.0 +2023-02-06 11:13:49,359 INFO [train.py:901] (1/4) Epoch 11, batch 8000, loss[loss=0.2388, simple_loss=0.3118, pruned_loss=0.08285, over 7967.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3103, pruned_loss=0.08036, over 1611234.92 frames. ], batch size: 21, lr: 6.81e-03, grad_scale: 16.0 +2023-02-06 11:13:50,906 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:14:02,005 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.663e+02 2.603e+02 3.071e+02 3.730e+02 8.421e+02, threshold=6.141e+02, percent-clipped=3.0 +2023-02-06 11:14:07,246 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88858.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:14:22,240 INFO [train.py:901] (1/4) Epoch 11, batch 8050, loss[loss=0.1961, simple_loss=0.2722, pruned_loss=0.06003, over 7215.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3084, pruned_loss=0.07986, over 1599023.27 frames. ], batch size: 16, lr: 6.81e-03, grad_scale: 16.0 +2023-02-06 11:14:54,022 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 11:14:58,683 INFO [train.py:901] (1/4) Epoch 12, batch 0, loss[loss=0.2134, simple_loss=0.2801, pruned_loss=0.07332, over 7699.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2801, pruned_loss=0.07332, over 7699.00 frames. ], batch size: 18, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:14:58,683 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 11:15:09,785 INFO [train.py:935] (1/4) Epoch 12, validation: loss=0.1897, simple_loss=0.2896, pruned_loss=0.04486, over 944034.00 frames. +2023-02-06 11:15:09,787 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 11:15:23,306 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 11:15:35,201 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.690e+02 3.540e+02 4.339e+02 7.249e+02, threshold=7.080e+02, percent-clipped=5.0 +2023-02-06 11:15:44,678 INFO [train.py:901] (1/4) Epoch 12, batch 50, loss[loss=0.1773, simple_loss=0.258, pruned_loss=0.04827, over 7213.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3154, pruned_loss=0.08223, over 366956.02 frames. ], batch size: 16, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:15:57,421 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 11:16:19,049 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 11:16:19,762 INFO [train.py:901] (1/4) Epoch 12, batch 100, loss[loss=0.1978, simple_loss=0.2901, pruned_loss=0.0527, over 8537.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3133, pruned_loss=0.08172, over 641890.29 frames. ], batch size: 28, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:16:26,519 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89024.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:16:33,362 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.57 vs. limit=5.0 +2023-02-06 11:16:40,643 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:16:43,463 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89049.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:16:43,905 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.547e+02 2.771e+02 3.256e+02 4.152e+02 1.357e+03, threshold=6.512e+02, percent-clipped=1.0 +2023-02-06 11:16:54,729 INFO [train.py:901] (1/4) Epoch 12, batch 150, loss[loss=0.1647, simple_loss=0.2372, pruned_loss=0.04612, over 7222.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3099, pruned_loss=0.07889, over 856206.81 frames. ], batch size: 16, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:17:24,726 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-06 11:17:29,011 INFO [train.py:901] (1/4) Epoch 12, batch 200, loss[loss=0.2347, simple_loss=0.3096, pruned_loss=0.07988, over 8490.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3097, pruned_loss=0.07866, over 1024675.60 frames. ], batch size: 28, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:17:53,942 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.699e+02 2.712e+02 3.423e+02 4.383e+02 1.008e+03, threshold=6.845e+02, percent-clipped=3.0 +2023-02-06 11:18:03,563 INFO [train.py:901] (1/4) Epoch 12, batch 250, loss[loss=0.1907, simple_loss=0.2813, pruned_loss=0.05004, over 8302.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3101, pruned_loss=0.07894, over 1155497.99 frames. ], batch size: 23, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:18:13,242 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 11:18:17,195 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.23 vs. limit=5.0 +2023-02-06 11:18:20,887 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89187.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:18:22,861 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 11:18:40,049 INFO [train.py:901] (1/4) Epoch 12, batch 300, loss[loss=0.2048, simple_loss=0.2773, pruned_loss=0.06614, over 7692.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3103, pruned_loss=0.07921, over 1260634.17 frames. ], batch size: 18, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:19:05,001 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.536e+02 3.052e+02 3.921e+02 6.584e+02, threshold=6.103e+02, percent-clipped=0.0 +2023-02-06 11:19:14,503 INFO [train.py:901] (1/4) Epoch 12, batch 350, loss[loss=0.2834, simple_loss=0.3469, pruned_loss=0.11, over 8424.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3103, pruned_loss=0.07915, over 1343046.37 frames. ], batch size: 49, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:19:49,366 INFO [train.py:901] (1/4) Epoch 12, batch 400, loss[loss=0.2202, simple_loss=0.297, pruned_loss=0.07167, over 8074.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3106, pruned_loss=0.07935, over 1404475.29 frames. ], batch size: 21, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:20:14,267 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.417e+02 2.965e+02 3.513e+02 5.511e+02, threshold=5.929e+02, percent-clipped=0.0 +2023-02-06 11:20:24,227 INFO [train.py:901] (1/4) Epoch 12, batch 450, loss[loss=0.212, simple_loss=0.2904, pruned_loss=0.06684, over 8137.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3106, pruned_loss=0.07951, over 1450528.40 frames. ], batch size: 22, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:20:25,042 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6904, 1.4666, 5.8051, 2.1503, 5.2045, 4.9196, 5.3759, 5.2040], + device='cuda:1'), covar=tensor([0.0437, 0.4551, 0.0297, 0.3153, 0.0879, 0.0664, 0.0425, 0.0481], + device='cuda:1'), in_proj_covar=tensor([0.0490, 0.0571, 0.0579, 0.0529, 0.0606, 0.0511, 0.0507, 0.0575], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:20:40,991 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:20:43,189 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:20:58,868 INFO [train.py:901] (1/4) Epoch 12, batch 500, loss[loss=0.1793, simple_loss=0.2537, pruned_loss=0.0525, over 7920.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3098, pruned_loss=0.07905, over 1485763.48 frames. ], batch size: 20, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:21:19,379 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89443.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:21:22,225 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5193, 1.7444, 2.8010, 1.3281, 2.0752, 1.8942, 1.6053, 1.9172], + device='cuda:1'), covar=tensor([0.1615, 0.1998, 0.0714, 0.3603, 0.1439, 0.2681, 0.1715, 0.1819], + device='cuda:1'), in_proj_covar=tensor([0.0485, 0.0521, 0.0534, 0.0580, 0.0619, 0.0555, 0.0473, 0.0612], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:21:24,107 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.539e+02 3.031e+02 3.696e+02 8.346e+02, threshold=6.063e+02, percent-clipped=3.0 +2023-02-06 11:21:29,672 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89457.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:21:34,341 INFO [train.py:901] (1/4) Epoch 12, batch 550, loss[loss=0.2208, simple_loss=0.2817, pruned_loss=0.07997, over 7659.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3106, pruned_loss=0.08025, over 1514428.92 frames. ], batch size: 19, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:21:53,122 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6006, 1.8986, 1.6653, 2.3168, 0.9657, 1.4475, 1.5524, 1.9149], + device='cuda:1'), covar=tensor([0.0906, 0.0801, 0.1041, 0.0435, 0.1220, 0.1410, 0.0891, 0.0756], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0213, 0.0255, 0.0214, 0.0218, 0.0251, 0.0257, 0.0220], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 11:22:02,583 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:22:09,198 INFO [train.py:901] (1/4) Epoch 12, batch 600, loss[loss=0.2304, simple_loss=0.3162, pruned_loss=0.0723, over 8516.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.311, pruned_loss=0.08026, over 1540257.39 frames. ], batch size: 26, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:22:17,833 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89527.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:22:20,461 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:22:26,550 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 11:22:34,513 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.630e+02 3.047e+02 3.733e+02 1.036e+03, threshold=6.094e+02, percent-clipped=2.0 +2023-02-06 11:22:44,045 INFO [train.py:901] (1/4) Epoch 12, batch 650, loss[loss=0.2783, simple_loss=0.3271, pruned_loss=0.1148, over 6820.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3105, pruned_loss=0.08023, over 1554058.65 frames. ], batch size: 15, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:23:18,870 INFO [train.py:901] (1/4) Epoch 12, batch 700, loss[loss=0.2593, simple_loss=0.3371, pruned_loss=0.09074, over 8335.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3103, pruned_loss=0.08008, over 1568926.12 frames. ], batch size: 25, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:23:40,433 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89646.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:23:43,631 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.821e+02 3.296e+02 4.031e+02 9.579e+02, threshold=6.593e+02, percent-clipped=5.0 +2023-02-06 11:23:47,395 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5678, 1.9930, 3.3348, 1.3328, 2.3409, 2.0170, 1.6828, 2.2380], + device='cuda:1'), covar=tensor([0.1636, 0.2257, 0.0636, 0.3928, 0.1658, 0.2828, 0.1755, 0.2251], + device='cuda:1'), in_proj_covar=tensor([0.0483, 0.0516, 0.0528, 0.0578, 0.0614, 0.0551, 0.0468, 0.0607], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:23:53,354 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9577, 1.6072, 1.3163, 1.5233, 1.2568, 1.1270, 1.1889, 1.2791], + device='cuda:1'), covar=tensor([0.1035, 0.0424, 0.1198, 0.0513, 0.0671, 0.1329, 0.0790, 0.0731], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0241, 0.0320, 0.0300, 0.0303, 0.0325, 0.0340, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 11:23:53,837 INFO [train.py:901] (1/4) Epoch 12, batch 750, loss[loss=0.261, simple_loss=0.3376, pruned_loss=0.09219, over 8341.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3108, pruned_loss=0.08077, over 1579016.42 frames. ], batch size: 26, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:23:55,356 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1441, 1.4110, 4.3126, 1.6009, 3.8456, 3.5940, 3.8986, 3.7342], + device='cuda:1'), covar=tensor([0.0506, 0.4121, 0.0457, 0.3188, 0.0927, 0.0830, 0.0504, 0.0605], + device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0576, 0.0582, 0.0530, 0.0606, 0.0516, 0.0509, 0.0579], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:24:02,992 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.41 vs. limit=5.0 +2023-02-06 11:24:11,504 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 11:24:17,637 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:24:20,260 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 11:24:28,200 INFO [train.py:901] (1/4) Epoch 12, batch 800, loss[loss=0.2491, simple_loss=0.3426, pruned_loss=0.07781, over 8094.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3113, pruned_loss=0.08047, over 1590170.64 frames. ], batch size: 23, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:24:37,596 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4430, 1.4204, 1.7115, 1.3417, 1.0015, 1.7216, 0.1737, 1.2148], + device='cuda:1'), covar=tensor([0.3014, 0.1917, 0.0546, 0.1757, 0.4207, 0.0584, 0.3429, 0.2129], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0171, 0.0102, 0.0218, 0.0259, 0.0107, 0.0166, 0.0169], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 11:24:43,681 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:24:53,399 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.628e+02 3.285e+02 4.121e+02 9.349e+02, threshold=6.571e+02, percent-clipped=6.0 +2023-02-06 11:24:59,656 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:25:02,846 INFO [train.py:901] (1/4) Epoch 12, batch 850, loss[loss=0.2178, simple_loss=0.2979, pruned_loss=0.06892, over 8501.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.311, pruned_loss=0.0806, over 1595393.51 frames. ], batch size: 31, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:25:18,000 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:25:19,284 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89787.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:25:28,670 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89801.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:25:31,255 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8388, 5.9855, 5.1686, 2.3850, 5.2506, 5.6058, 5.3778, 5.3142], + device='cuda:1'), covar=tensor([0.0681, 0.0442, 0.0872, 0.4594, 0.0758, 0.0700, 0.1114, 0.0591], + device='cuda:1'), in_proj_covar=tensor([0.0473, 0.0383, 0.0388, 0.0491, 0.0390, 0.0386, 0.0380, 0.0338], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:25:37,800 INFO [train.py:901] (1/4) Epoch 12, batch 900, loss[loss=0.2032, simple_loss=0.2869, pruned_loss=0.05972, over 8191.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3102, pruned_loss=0.07931, over 1605862.91 frames. ], batch size: 23, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:26:03,293 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.446e+02 3.021e+02 3.729e+02 6.397e+02, threshold=6.041e+02, percent-clipped=0.0 +2023-02-06 11:26:03,491 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:26:11,263 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6119, 1.5050, 5.6521, 2.2671, 5.0213, 4.7954, 5.2760, 5.0402], + device='cuda:1'), covar=tensor([0.0401, 0.4638, 0.0364, 0.3309, 0.0950, 0.0660, 0.0390, 0.0485], + device='cuda:1'), in_proj_covar=tensor([0.0500, 0.0586, 0.0593, 0.0542, 0.0619, 0.0526, 0.0519, 0.0593], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:26:11,800 INFO [train.py:901] (1/4) Epoch 12, batch 950, loss[loss=0.2501, simple_loss=0.3303, pruned_loss=0.08497, over 8524.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3101, pruned_loss=0.07914, over 1605389.37 frames. ], batch size: 34, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:26:16,404 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89871.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:26:24,520 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89883.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:26:30,578 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 11:26:38,567 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89902.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:26:38,607 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89902.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:26:39,096 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 11:26:46,288 INFO [train.py:901] (1/4) Epoch 12, batch 1000, loss[loss=0.2793, simple_loss=0.3408, pruned_loss=0.1089, over 8190.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.311, pruned_loss=0.07989, over 1607141.80 frames. ], batch size: 23, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:26:47,815 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89916.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:26:55,678 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:27:11,391 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.648e+02 3.254e+02 4.081e+02 9.414e+02, threshold=6.507e+02, percent-clipped=7.0 +2023-02-06 11:27:11,417 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 11:27:20,829 INFO [train.py:901] (1/4) Epoch 12, batch 1050, loss[loss=0.2308, simple_loss=0.3067, pruned_loss=0.07748, over 8467.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3105, pruned_loss=0.07955, over 1609810.85 frames. ], batch size: 25, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:27:24,329 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 11:27:35,822 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89986.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:27:56,236 INFO [train.py:901] (1/4) Epoch 12, batch 1100, loss[loss=0.1973, simple_loss=0.285, pruned_loss=0.05476, over 7926.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3108, pruned_loss=0.07989, over 1609214.09 frames. ], batch size: 20, lr: 6.48e-03, grad_scale: 8.0 +2023-02-06 11:27:57,787 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4820, 1.2904, 2.3608, 1.1276, 2.0695, 2.4882, 2.6235, 2.1451], + device='cuda:1'), covar=tensor([0.0871, 0.1173, 0.0429, 0.1922, 0.0717, 0.0356, 0.0561, 0.0686], + device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0296, 0.0260, 0.0292, 0.0274, 0.0236, 0.0345, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 11:28:15,782 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:28:22,906 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.536e+02 3.046e+02 3.976e+02 6.882e+02, threshold=6.092e+02, percent-clipped=1.0 +2023-02-06 11:28:31,018 INFO [train.py:901] (1/4) Epoch 12, batch 1150, loss[loss=0.2447, simple_loss=0.3297, pruned_loss=0.07991, over 8496.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3092, pruned_loss=0.07907, over 1607020.57 frames. ], batch size: 26, lr: 6.48e-03, grad_scale: 4.0 +2023-02-06 11:28:34,440 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 11:29:01,090 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:29:05,597 INFO [train.py:901] (1/4) Epoch 12, batch 1200, loss[loss=0.2789, simple_loss=0.3405, pruned_loss=0.1086, over 8187.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3103, pruned_loss=0.07955, over 1611330.87 frames. ], batch size: 23, lr: 6.48e-03, grad_scale: 8.0 +2023-02-06 11:29:19,573 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90132.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:29:30,361 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90148.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:29:32,957 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.449e+02 3.099e+02 4.282e+02 6.791e+02, threshold=6.197e+02, percent-clipped=4.0 +2023-02-06 11:29:36,548 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90157.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:29:37,309 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90158.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:29:41,828 INFO [train.py:901] (1/4) Epoch 12, batch 1250, loss[loss=0.1873, simple_loss=0.263, pruned_loss=0.05579, over 7814.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3101, pruned_loss=0.07909, over 1615725.54 frames. ], batch size: 20, lr: 6.48e-03, grad_scale: 8.0 +2023-02-06 11:29:47,517 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90172.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:29:55,698 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90183.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:30:05,420 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90197.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:30:17,140 INFO [train.py:901] (1/4) Epoch 12, batch 1300, loss[loss=0.2213, simple_loss=0.3009, pruned_loss=0.07086, over 8198.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3103, pruned_loss=0.07886, over 1615961.42 frames. ], batch size: 23, lr: 6.48e-03, grad_scale: 8.0 +2023-02-06 11:30:26,149 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90227.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:30:37,504 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90242.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:30:43,679 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:30:44,948 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.408e+02 3.209e+02 4.069e+02 1.568e+03, threshold=6.418e+02, percent-clipped=9.0 +2023-02-06 11:30:53,231 INFO [train.py:901] (1/4) Epoch 12, batch 1350, loss[loss=0.3019, simple_loss=0.3525, pruned_loss=0.1256, over 6659.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3101, pruned_loss=0.07853, over 1613379.83 frames. ], batch size: 71, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:30:55,580 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90267.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:31:09,750 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0848, 1.6196, 1.3630, 1.6333, 1.3599, 1.2188, 1.2858, 1.3874], + device='cuda:1'), covar=tensor([0.1007, 0.0426, 0.1203, 0.0465, 0.0601, 0.1307, 0.0795, 0.0655], + device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0240, 0.0318, 0.0299, 0.0301, 0.0325, 0.0339, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 11:31:28,597 INFO [train.py:901] (1/4) Epoch 12, batch 1400, loss[loss=0.2679, simple_loss=0.3451, pruned_loss=0.0954, over 8287.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3101, pruned_loss=0.0792, over 1615178.30 frames. ], batch size: 23, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:31:30,297 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-06 11:31:47,887 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:31:54,627 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.398e+02 2.808e+02 3.540e+02 8.131e+02, threshold=5.617e+02, percent-clipped=1.0 +2023-02-06 11:32:03,614 INFO [train.py:901] (1/4) Epoch 12, batch 1450, loss[loss=0.2806, simple_loss=0.3415, pruned_loss=0.1098, over 8107.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3094, pruned_loss=0.07905, over 1613269.05 frames. ], batch size: 23, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:32:07,898 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90369.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:32:08,434 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 11:32:38,144 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90413.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:32:38,593 INFO [train.py:901] (1/4) Epoch 12, batch 1500, loss[loss=0.2122, simple_loss=0.2917, pruned_loss=0.06636, over 8284.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3081, pruned_loss=0.079, over 1612834.48 frames. ], batch size: 23, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:32:55,147 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:33:04,320 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 2.462e+02 2.993e+02 3.898e+02 9.256e+02, threshold=5.985e+02, percent-clipped=2.0 +2023-02-06 11:33:12,495 INFO [train.py:901] (1/4) Epoch 12, batch 1550, loss[loss=0.2469, simple_loss=0.3134, pruned_loss=0.09022, over 8133.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.309, pruned_loss=0.07888, over 1617082.15 frames. ], batch size: 22, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:33:21,467 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:33:33,069 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90492.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:33:48,245 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0483, 2.4242, 1.7789, 2.7813, 1.5337, 1.5150, 2.0724, 2.3359], + device='cuda:1'), covar=tensor([0.0721, 0.0821, 0.1057, 0.0389, 0.1139, 0.1407, 0.0924, 0.0831], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0215, 0.0258, 0.0218, 0.0221, 0.0252, 0.0259, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 11:33:48,745 INFO [train.py:901] (1/4) Epoch 12, batch 1600, loss[loss=0.252, simple_loss=0.3383, pruned_loss=0.08289, over 8513.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3088, pruned_loss=0.0781, over 1620417.03 frames. ], batch size: 26, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:33:48,905 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90514.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:34:15,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.379e+02 2.712e+02 3.378e+02 4.197e+02 8.231e+02, threshold=6.755e+02, percent-clipped=6.0 +2023-02-06 11:34:23,538 INFO [train.py:901] (1/4) Epoch 12, batch 1650, loss[loss=0.2163, simple_loss=0.3016, pruned_loss=0.06545, over 7974.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.308, pruned_loss=0.07781, over 1618711.71 frames. ], batch size: 21, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:34:43,461 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:34:44,973 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7148, 1.3259, 1.5904, 1.2361, 0.9727, 1.3753, 1.4640, 1.3472], + device='cuda:1'), covar=tensor([0.0467, 0.1224, 0.1636, 0.1378, 0.0540, 0.1470, 0.0685, 0.0620], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0152, 0.0191, 0.0158, 0.0103, 0.0163, 0.0117, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 11:34:47,040 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90598.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:34:53,818 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:34:58,208 INFO [train.py:901] (1/4) Epoch 12, batch 1700, loss[loss=0.3033, simple_loss=0.3585, pruned_loss=0.1241, over 8035.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3089, pruned_loss=0.07862, over 1617882.45 frames. ], batch size: 22, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:35:04,319 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:35:12,050 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.16 vs. limit=2.0 +2023-02-06 11:35:18,595 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3835, 2.0208, 3.4044, 1.2617, 2.5182, 1.9459, 1.5919, 2.4185], + device='cuda:1'), covar=tensor([0.1749, 0.2115, 0.0667, 0.3885, 0.1419, 0.2782, 0.1834, 0.2076], + device='cuda:1'), in_proj_covar=tensor([0.0484, 0.0518, 0.0533, 0.0581, 0.0617, 0.0556, 0.0470, 0.0614], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:35:24,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.486e+02 2.952e+02 3.646e+02 6.764e+02, threshold=5.904e+02, percent-clipped=1.0 +2023-02-06 11:35:33,331 INFO [train.py:901] (1/4) Epoch 12, batch 1750, loss[loss=0.206, simple_loss=0.2934, pruned_loss=0.05931, over 7975.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3084, pruned_loss=0.07827, over 1612624.29 frames. ], batch size: 21, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:36:04,022 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:36:07,404 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90713.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:36:08,032 INFO [train.py:901] (1/4) Epoch 12, batch 1800, loss[loss=0.2485, simple_loss=0.3135, pruned_loss=0.09178, over 7703.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3084, pruned_loss=0.07821, over 1614360.51 frames. ], batch size: 18, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:36:35,308 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.841e+02 2.625e+02 3.119e+02 3.569e+02 7.012e+02, threshold=6.239e+02, percent-clipped=2.0 +2023-02-06 11:36:43,320 INFO [train.py:901] (1/4) Epoch 12, batch 1850, loss[loss=0.2074, simple_loss=0.2876, pruned_loss=0.06358, over 8291.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3091, pruned_loss=0.07872, over 1612417.96 frames. ], batch size: 23, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:37:17,706 INFO [train.py:901] (1/4) Epoch 12, batch 1900, loss[loss=0.2539, simple_loss=0.3342, pruned_loss=0.08675, over 8022.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3096, pruned_loss=0.07887, over 1613642.41 frames. ], batch size: 22, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:37:22,469 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90821.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:27,322 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90828.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:44,438 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.569e+02 3.031e+02 3.632e+02 7.649e+02, threshold=6.063e+02, percent-clipped=2.0 +2023-02-06 11:37:47,238 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 11:37:48,679 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90858.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:52,154 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90863.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:52,632 INFO [train.py:901] (1/4) Epoch 12, batch 1950, loss[loss=0.2084, simple_loss=0.2873, pruned_loss=0.06477, over 8037.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3105, pruned_loss=0.07976, over 1618232.53 frames. ], batch size: 22, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:37:54,799 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90867.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:59,342 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 11:38:10,331 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:38:19,030 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 11:38:27,972 INFO [train.py:901] (1/4) Epoch 12, batch 2000, loss[loss=0.2301, simple_loss=0.2995, pruned_loss=0.08037, over 8275.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.31, pruned_loss=0.0792, over 1619907.51 frames. ], batch size: 23, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:38:36,468 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9460, 2.3655, 3.6582, 1.7560, 3.0283, 2.2970, 2.0825, 2.7035], + device='cuda:1'), covar=tensor([0.1285, 0.1873, 0.0528, 0.3127, 0.1112, 0.2384, 0.1420, 0.1892], + device='cuda:1'), in_proj_covar=tensor([0.0485, 0.0518, 0.0531, 0.0582, 0.0616, 0.0558, 0.0471, 0.0610], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:38:43,405 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:38:54,918 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 2.641e+02 3.163e+02 4.034e+02 9.087e+02, threshold=6.326e+02, percent-clipped=9.0 +2023-02-06 11:39:02,897 INFO [train.py:901] (1/4) Epoch 12, batch 2050, loss[loss=0.2459, simple_loss=0.3099, pruned_loss=0.09094, over 7917.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3078, pruned_loss=0.07805, over 1614766.51 frames. ], batch size: 20, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:39:03,755 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:39:09,924 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90973.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:39:21,820 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90990.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:39:38,681 INFO [train.py:901] (1/4) Epoch 12, batch 2100, loss[loss=0.1899, simple_loss=0.2707, pruned_loss=0.05457, over 8079.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3086, pruned_loss=0.07808, over 1617448.58 frames. ], batch size: 21, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:39:40,484 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 11:40:04,173 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.877e+02 2.659e+02 3.265e+02 4.247e+02 8.349e+02, threshold=6.531e+02, percent-clipped=2.0 +2023-02-06 11:40:12,099 INFO [train.py:901] (1/4) Epoch 12, batch 2150, loss[loss=0.233, simple_loss=0.3092, pruned_loss=0.07842, over 8252.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3079, pruned_loss=0.07771, over 1615672.79 frames. ], batch size: 24, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:40:26,821 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91084.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:40:44,037 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91109.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:40:47,215 INFO [train.py:901] (1/4) Epoch 12, batch 2200, loss[loss=0.2332, simple_loss=0.2998, pruned_loss=0.08333, over 8036.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3081, pruned_loss=0.0783, over 1614756.14 frames. ], batch size: 22, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:41:13,716 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.817e+02 2.751e+02 3.546e+02 4.173e+02 9.054e+02, threshold=7.092e+02, percent-clipped=3.0 +2023-02-06 11:41:21,758 INFO [train.py:901] (1/4) Epoch 12, batch 2250, loss[loss=0.2348, simple_loss=0.2955, pruned_loss=0.08703, over 7780.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3081, pruned_loss=0.07819, over 1614347.04 frames. ], batch size: 19, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:41:41,116 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91192.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:41:44,211 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 11:41:54,520 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91211.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:41:56,402 INFO [train.py:901] (1/4) Epoch 12, batch 2300, loss[loss=0.2192, simple_loss=0.2857, pruned_loss=0.07638, over 7264.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3077, pruned_loss=0.07846, over 1611356.54 frames. ], batch size: 16, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:41:58,490 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91217.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:42:01,936 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0203, 1.5599, 1.6486, 1.4791, 0.9079, 1.4666, 1.6799, 1.6076], + device='cuda:1'), covar=tensor([0.0474, 0.1215, 0.1591, 0.1303, 0.0598, 0.1427, 0.0655, 0.0566], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0152, 0.0190, 0.0159, 0.0103, 0.0162, 0.0115, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 11:42:07,343 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91229.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:42:15,377 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5300, 2.6677, 1.9321, 2.2585, 2.2573, 1.4640, 2.0612, 2.2783], + device='cuda:1'), covar=tensor([0.1528, 0.0374, 0.1068, 0.0696, 0.0692, 0.1590, 0.0984, 0.0843], + device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0237, 0.0318, 0.0295, 0.0300, 0.0325, 0.0341, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 11:42:23,423 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.635e+02 3.142e+02 4.194e+02 9.102e+02, threshold=6.284e+02, percent-clipped=2.0 +2023-02-06 11:42:25,004 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91254.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:42:31,176 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3506, 1.5339, 1.3847, 1.9048, 0.7307, 1.2204, 1.3210, 1.5577], + device='cuda:1'), covar=tensor([0.0896, 0.0795, 0.1180, 0.0524, 0.1220, 0.1378, 0.0792, 0.0839], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0211, 0.0256, 0.0215, 0.0218, 0.0253, 0.0258, 0.0220], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 11:42:31,680 INFO [train.py:901] (1/4) Epoch 12, batch 2350, loss[loss=0.2391, simple_loss=0.3003, pruned_loss=0.08896, over 7439.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3077, pruned_loss=0.07826, over 1614908.14 frames. ], batch size: 17, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:42:40,965 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8013, 5.8127, 5.0339, 2.4314, 5.1398, 5.4676, 5.3685, 5.2020], + device='cuda:1'), covar=tensor([0.0490, 0.0364, 0.0830, 0.4558, 0.0601, 0.0677, 0.1023, 0.0508], + device='cuda:1'), in_proj_covar=tensor([0.0462, 0.0372, 0.0385, 0.0484, 0.0378, 0.0382, 0.0377, 0.0330], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:42:47,155 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3348, 2.0766, 1.6543, 1.9612, 1.7655, 1.3697, 1.6356, 1.7983], + device='cuda:1'), covar=tensor([0.1247, 0.0386, 0.1093, 0.0550, 0.0658, 0.1405, 0.0870, 0.0756], + device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0237, 0.0318, 0.0296, 0.0300, 0.0325, 0.0342, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 11:42:49,223 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4188, 1.8083, 3.1015, 1.1517, 2.0999, 1.6844, 1.5192, 1.8565], + device='cuda:1'), covar=tensor([0.1729, 0.2085, 0.0680, 0.3925, 0.1656, 0.2938, 0.1801, 0.2380], + device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0521, 0.0534, 0.0583, 0.0619, 0.0558, 0.0471, 0.0612], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:42:57,738 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91303.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:43:02,771 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 11:43:05,258 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6218, 2.2693, 4.4292, 1.3595, 3.0167, 2.2245, 1.8636, 2.8649], + device='cuda:1'), covar=tensor([0.1709, 0.2251, 0.0629, 0.3945, 0.1563, 0.2761, 0.1716, 0.2194], + device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0522, 0.0535, 0.0582, 0.0619, 0.0557, 0.0471, 0.0611], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:43:05,699 INFO [train.py:901] (1/4) Epoch 12, batch 2400, loss[loss=0.2184, simple_loss=0.2996, pruned_loss=0.06859, over 8292.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3076, pruned_loss=0.07835, over 1613126.74 frames. ], batch size: 23, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:43:14,307 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91326.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:43:32,237 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.547e+02 3.046e+02 3.774e+02 7.420e+02, threshold=6.092e+02, percent-clipped=3.0 +2023-02-06 11:43:41,055 INFO [train.py:901] (1/4) Epoch 12, batch 2450, loss[loss=0.1812, simple_loss=0.2557, pruned_loss=0.05338, over 7696.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3079, pruned_loss=0.07875, over 1614351.62 frames. ], batch size: 18, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:44:15,047 INFO [train.py:901] (1/4) Epoch 12, batch 2500, loss[loss=0.2616, simple_loss=0.3321, pruned_loss=0.09556, over 8465.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3084, pruned_loss=0.07907, over 1616305.17 frames. ], batch size: 25, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:44:27,373 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4565, 1.8157, 4.5257, 1.7465, 2.6152, 5.0486, 5.1747, 3.9649], + device='cuda:1'), covar=tensor([0.1194, 0.1734, 0.0275, 0.2339, 0.1010, 0.0268, 0.0371, 0.0891], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0301, 0.0264, 0.0296, 0.0280, 0.0237, 0.0351, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 11:44:41,745 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.576e+02 3.186e+02 4.386e+02 8.083e+02, threshold=6.372e+02, percent-clipped=11.0 +2023-02-06 11:44:50,281 INFO [train.py:901] (1/4) Epoch 12, batch 2550, loss[loss=0.2358, simple_loss=0.3153, pruned_loss=0.07819, over 8446.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3087, pruned_loss=0.07928, over 1618227.98 frames. ], batch size: 27, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:45:24,410 INFO [train.py:901] (1/4) Epoch 12, batch 2600, loss[loss=0.2012, simple_loss=0.2652, pruned_loss=0.06861, over 7460.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3088, pruned_loss=0.07984, over 1615039.70 frames. ], batch size: 17, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:45:45,627 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4412, 2.0499, 3.4149, 1.2597, 2.4983, 1.8312, 1.5871, 2.3689], + device='cuda:1'), covar=tensor([0.1773, 0.2122, 0.0746, 0.3952, 0.1646, 0.2903, 0.1804, 0.2237], + device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0523, 0.0536, 0.0584, 0.0620, 0.0556, 0.0472, 0.0614], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:45:50,013 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.880e+02 3.430e+02 4.544e+02 8.443e+02, threshold=6.860e+02, percent-clipped=9.0 +2023-02-06 11:45:56,362 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91560.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:45:58,895 INFO [train.py:901] (1/4) Epoch 12, batch 2650, loss[loss=0.2084, simple_loss=0.2844, pruned_loss=0.06618, over 8141.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3089, pruned_loss=0.07961, over 1615214.59 frames. ], batch size: 22, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:46:11,917 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91582.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:46:21,317 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6334, 2.1718, 4.4158, 1.3442, 3.0710, 2.1457, 1.6727, 2.9688], + device='cuda:1'), covar=tensor([0.1687, 0.2374, 0.0519, 0.4080, 0.1494, 0.2850, 0.1871, 0.2140], + device='cuda:1'), in_proj_covar=tensor([0.0488, 0.0527, 0.0539, 0.0587, 0.0623, 0.0560, 0.0475, 0.0617], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:46:29,372 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:46:33,855 INFO [train.py:901] (1/4) Epoch 12, batch 2700, loss[loss=0.2043, simple_loss=0.2877, pruned_loss=0.06049, over 7962.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3097, pruned_loss=0.08016, over 1611814.11 frames. ], batch size: 21, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:46:42,003 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5184, 1.9291, 3.0644, 1.2980, 2.1764, 1.8184, 1.6622, 2.0222], + device='cuda:1'), covar=tensor([0.1610, 0.1946, 0.0689, 0.3729, 0.1501, 0.2724, 0.1627, 0.2004], + device='cuda:1'), in_proj_covar=tensor([0.0485, 0.0523, 0.0536, 0.0583, 0.0617, 0.0558, 0.0472, 0.0612], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:46:55,931 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91647.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:46:59,290 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.691e+02 3.205e+02 3.908e+02 7.628e+02, threshold=6.410e+02, percent-clipped=2.0 +2023-02-06 11:47:08,031 INFO [train.py:901] (1/4) Epoch 12, batch 2750, loss[loss=0.1961, simple_loss=0.2846, pruned_loss=0.05379, over 8317.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3079, pruned_loss=0.0785, over 1612792.03 frames. ], batch size: 25, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:47:10,606 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3812, 2.6720, 2.1354, 3.7673, 1.5955, 1.8859, 2.1536, 3.0557], + device='cuda:1'), covar=tensor([0.0753, 0.1004, 0.1050, 0.0337, 0.1351, 0.1470, 0.1329, 0.0768], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0214, 0.0258, 0.0220, 0.0220, 0.0256, 0.0262, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 11:47:23,579 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4851, 1.4328, 1.7892, 1.4559, 1.1008, 1.8229, 0.0936, 1.1590], + device='cuda:1'), covar=tensor([0.2422, 0.1935, 0.0513, 0.1291, 0.3982, 0.0548, 0.3309, 0.1754], + device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0174, 0.0103, 0.0218, 0.0258, 0.0107, 0.0167, 0.0168], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 11:47:27,141 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 11:47:38,296 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.2471, 5.1426, 4.6023, 2.5089, 4.6308, 4.8572, 4.8988, 4.5255], + device='cuda:1'), covar=tensor([0.0489, 0.0399, 0.0842, 0.4008, 0.0638, 0.0841, 0.1021, 0.0663], + device='cuda:1'), in_proj_covar=tensor([0.0465, 0.0377, 0.0389, 0.0486, 0.0380, 0.0386, 0.0375, 0.0332], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:47:43,511 INFO [train.py:901] (1/4) Epoch 12, batch 2800, loss[loss=0.211, simple_loss=0.288, pruned_loss=0.06701, over 7214.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3082, pruned_loss=0.07859, over 1615055.33 frames. ], batch size: 16, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:48:08,847 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.634e+02 3.181e+02 3.784e+02 9.192e+02, threshold=6.362e+02, percent-clipped=3.0 +2023-02-06 11:48:15,802 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91762.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:48:16,915 INFO [train.py:901] (1/4) Epoch 12, batch 2850, loss[loss=0.2634, simple_loss=0.3141, pruned_loss=0.1063, over 7532.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3094, pruned_loss=0.07938, over 1615185.85 frames. ], batch size: 18, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:48:48,409 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3447, 2.4060, 1.9234, 2.9503, 1.5659, 1.7589, 2.2090, 2.5242], + device='cuda:1'), covar=tensor([0.0577, 0.0721, 0.0883, 0.0355, 0.1020, 0.1208, 0.0844, 0.0608], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0214, 0.0258, 0.0219, 0.0218, 0.0254, 0.0260, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 11:48:52,944 INFO [train.py:901] (1/4) Epoch 12, batch 2900, loss[loss=0.2117, simple_loss=0.2762, pruned_loss=0.07361, over 7698.00 frames. ], tot_loss[loss=0.2328, simple_loss=0.3082, pruned_loss=0.07874, over 1610253.00 frames. ], batch size: 18, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:49:18,828 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.538e+02 3.175e+02 3.875e+02 8.885e+02, threshold=6.349e+02, percent-clipped=4.0 +2023-02-06 11:49:22,156 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 11:49:26,762 INFO [train.py:901] (1/4) Epoch 12, batch 2950, loss[loss=0.234, simple_loss=0.3071, pruned_loss=0.08046, over 8334.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3084, pruned_loss=0.07883, over 1608226.13 frames. ], batch size: 26, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:49:54,042 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91904.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:49:57,393 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91909.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:50:00,613 INFO [train.py:901] (1/4) Epoch 12, batch 3000, loss[loss=0.2033, simple_loss=0.283, pruned_loss=0.06177, over 8231.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3092, pruned_loss=0.07956, over 1608342.14 frames. ], batch size: 22, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:50:00,613 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 11:50:13,630 INFO [train.py:935] (1/4) Epoch 12, validation: loss=0.1868, simple_loss=0.2871, pruned_loss=0.04323, over 944034.00 frames. +2023-02-06 11:50:13,632 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6608MB +2023-02-06 11:50:26,349 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0205, 3.0736, 2.8143, 4.1868, 1.8019, 2.5107, 2.6129, 3.5395], + device='cuda:1'), covar=tensor([0.0610, 0.0874, 0.0823, 0.0270, 0.1253, 0.1218, 0.1147, 0.0727], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0212, 0.0259, 0.0218, 0.0219, 0.0253, 0.0260, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 11:50:40,664 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.361e+02 2.883e+02 3.802e+02 7.578e+02, threshold=5.767e+02, percent-clipped=3.0 +2023-02-06 11:50:49,085 INFO [train.py:901] (1/4) Epoch 12, batch 3050, loss[loss=0.2228, simple_loss=0.2867, pruned_loss=0.07942, over 7443.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3089, pruned_loss=0.07934, over 1608248.70 frames. ], batch size: 17, lr: 6.41e-03, grad_scale: 8.0 +2023-02-06 11:50:56,132 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91973.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:51:14,066 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91999.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:51:25,121 INFO [train.py:901] (1/4) Epoch 12, batch 3100, loss[loss=0.2119, simple_loss=0.2984, pruned_loss=0.0627, over 8604.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3092, pruned_loss=0.07933, over 1610112.86 frames. ], batch size: 31, lr: 6.41e-03, grad_scale: 8.0 +2023-02-06 11:51:28,126 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:51:28,811 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92019.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:51:45,784 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:51:51,726 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.663e+02 3.347e+02 4.142e+02 7.838e+02, threshold=6.695e+02, percent-clipped=5.0 +2023-02-06 11:52:01,130 INFO [train.py:901] (1/4) Epoch 12, batch 3150, loss[loss=0.229, simple_loss=0.2989, pruned_loss=0.07955, over 8656.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3098, pruned_loss=0.07985, over 1615297.10 frames. ], batch size: 34, lr: 6.41e-03, grad_scale: 16.0 +2023-02-06 11:52:26,881 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 11:52:35,761 INFO [train.py:901] (1/4) Epoch 12, batch 3200, loss[loss=0.1841, simple_loss=0.2596, pruned_loss=0.05428, over 6854.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3096, pruned_loss=0.07925, over 1612438.85 frames. ], batch size: 15, lr: 6.41e-03, grad_scale: 16.0 +2023-02-06 11:53:02,009 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.674e+02 3.226e+02 3.971e+02 7.397e+02, threshold=6.453e+02, percent-clipped=3.0 +2023-02-06 11:53:10,368 INFO [train.py:901] (1/4) Epoch 12, batch 3250, loss[loss=0.2336, simple_loss=0.3034, pruned_loss=0.08187, over 8088.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3086, pruned_loss=0.07831, over 1618741.96 frames. ], batch size: 21, lr: 6.41e-03, grad_scale: 16.0 +2023-02-06 11:53:16,098 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1954, 1.6798, 1.8085, 1.5498, 1.1105, 1.6891, 1.8366, 1.9595], + device='cuda:1'), covar=tensor([0.0446, 0.1148, 0.1563, 0.1245, 0.0550, 0.1366, 0.0634, 0.0512], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0152, 0.0191, 0.0158, 0.0102, 0.0162, 0.0115, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:1') +2023-02-06 11:53:16,528 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 11:53:46,146 INFO [train.py:901] (1/4) Epoch 12, batch 3300, loss[loss=0.1989, simple_loss=0.2777, pruned_loss=0.06005, over 7980.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3076, pruned_loss=0.07755, over 1615092.51 frames. ], batch size: 21, lr: 6.41e-03, grad_scale: 16.0 +2023-02-06 11:53:53,631 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2961, 1.5420, 4.5307, 1.6924, 3.9269, 3.8209, 4.0785, 3.9509], + device='cuda:1'), covar=tensor([0.0583, 0.3779, 0.0504, 0.3231, 0.1164, 0.0866, 0.0617, 0.0708], + device='cuda:1'), in_proj_covar=tensor([0.0501, 0.0579, 0.0593, 0.0540, 0.0622, 0.0531, 0.0522, 0.0586], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:54:11,040 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.357e+02 2.935e+02 3.680e+02 6.719e+02, threshold=5.870e+02, percent-clipped=1.0 +2023-02-06 11:54:11,768 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92253.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:54:19,108 INFO [train.py:901] (1/4) Epoch 12, batch 3350, loss[loss=0.1947, simple_loss=0.2661, pruned_loss=0.0616, over 7717.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3087, pruned_loss=0.07872, over 1618785.66 frames. ], batch size: 18, lr: 6.40e-03, grad_scale: 16.0 +2023-02-06 11:54:27,349 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92275.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:54:45,174 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92300.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:54:55,326 INFO [train.py:901] (1/4) Epoch 12, batch 3400, loss[loss=0.3059, simple_loss=0.3529, pruned_loss=0.1294, over 8200.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3093, pruned_loss=0.07884, over 1618163.22 frames. ], batch size: 23, lr: 6.40e-03, grad_scale: 16.0 +2023-02-06 11:54:57,464 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92317.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:55:15,991 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92343.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:55:21,866 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.364e+02 2.893e+02 3.659e+02 6.777e+02, threshold=5.785e+02, percent-clipped=2.0 +2023-02-06 11:55:29,927 INFO [train.py:901] (1/4) Epoch 12, batch 3450, loss[loss=0.2176, simple_loss=0.3024, pruned_loss=0.06636, over 8468.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3093, pruned_loss=0.0786, over 1622022.61 frames. ], batch size: 25, lr: 6.40e-03, grad_scale: 16.0 +2023-02-06 11:55:32,847 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92368.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:55:39,753 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4623, 1.8184, 1.8856, 0.9838, 1.9504, 1.3367, 0.4465, 1.7184], + device='cuda:1'), covar=tensor([0.0401, 0.0230, 0.0177, 0.0406, 0.0251, 0.0634, 0.0615, 0.0178], + device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0334, 0.0286, 0.0397, 0.0326, 0.0486, 0.0364, 0.0365], + device='cuda:1'), out_proj_covar=tensor([1.1477e-04, 9.2415e-05, 7.9310e-05, 1.1088e-04, 9.1650e-05, 1.4679e-04, + 1.0322e-04, 1.0268e-04], device='cuda:1') +2023-02-06 11:56:00,344 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2186, 1.0419, 1.2739, 1.0967, 0.9835, 1.3039, 0.0390, 0.8691], + device='cuda:1'), covar=tensor([0.2325, 0.1928, 0.0615, 0.1198, 0.3560, 0.0654, 0.3093, 0.1839], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0174, 0.0104, 0.0219, 0.0256, 0.0107, 0.0165, 0.0167], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 11:56:04,889 INFO [train.py:901] (1/4) Epoch 12, batch 3500, loss[loss=0.2175, simple_loss=0.2922, pruned_loss=0.07137, over 7780.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3081, pruned_loss=0.07816, over 1619530.11 frames. ], batch size: 19, lr: 6.40e-03, grad_scale: 8.0 +2023-02-06 11:56:18,393 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92432.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:56:27,916 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7823, 1.5050, 3.9316, 1.3207, 3.4443, 3.2686, 3.5608, 3.4175], + device='cuda:1'), covar=tensor([0.0620, 0.3981, 0.0581, 0.3648, 0.1241, 0.0900, 0.0629, 0.0745], + device='cuda:1'), in_proj_covar=tensor([0.0498, 0.0574, 0.0595, 0.0538, 0.0621, 0.0532, 0.0522, 0.0586], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:56:29,149 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 11:56:33,046 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.547e+02 2.438e+02 2.928e+02 3.742e+02 8.211e+02, threshold=5.856e+02, percent-clipped=5.0 +2023-02-06 11:56:36,508 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92458.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:56:38,508 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4983, 1.8488, 1.8588, 1.0180, 1.9531, 1.3800, 0.4345, 1.7038], + device='cuda:1'), covar=tensor([0.0337, 0.0217, 0.0198, 0.0388, 0.0251, 0.0685, 0.0568, 0.0176], + device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0330, 0.0281, 0.0392, 0.0323, 0.0480, 0.0359, 0.0362], + device='cuda:1'), out_proj_covar=tensor([1.1321e-04, 9.1216e-05, 7.7866e-05, 1.0948e-04, 9.0566e-05, 1.4500e-04, + 1.0185e-04, 1.0176e-04], device='cuda:1') +2023-02-06 11:56:40,253 INFO [train.py:901] (1/4) Epoch 12, batch 3550, loss[loss=0.2118, simple_loss=0.3051, pruned_loss=0.05927, over 8097.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3082, pruned_loss=0.07796, over 1622270.60 frames. ], batch size: 23, lr: 6.40e-03, grad_scale: 8.0 +2023-02-06 11:57:11,660 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-02-06 11:57:14,698 INFO [train.py:901] (1/4) Epoch 12, batch 3600, loss[loss=0.2331, simple_loss=0.3001, pruned_loss=0.08303, over 5157.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3075, pruned_loss=0.07791, over 1616035.28 frames. ], batch size: 11, lr: 6.40e-03, grad_scale: 8.0 +2023-02-06 11:57:29,921 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.92 vs. limit=5.0 +2023-02-06 11:57:38,949 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.87 vs. limit=5.0 +2023-02-06 11:57:42,454 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 2.628e+02 3.055e+02 4.234e+02 9.851e+02, threshold=6.109e+02, percent-clipped=7.0 +2023-02-06 11:57:50,882 INFO [train.py:901] (1/4) Epoch 12, batch 3650, loss[loss=0.2167, simple_loss=0.2985, pruned_loss=0.06748, over 8237.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3065, pruned_loss=0.07738, over 1609207.98 frames. ], batch size: 22, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 11:58:10,860 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6876, 4.6987, 4.1192, 2.1471, 4.1725, 4.2618, 4.2554, 4.0610], + device='cuda:1'), covar=tensor([0.0711, 0.0517, 0.1076, 0.5019, 0.0729, 0.0993, 0.1291, 0.0771], + device='cuda:1'), in_proj_covar=tensor([0.0468, 0.0377, 0.0388, 0.0488, 0.0380, 0.0383, 0.0379, 0.0332], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 11:58:23,907 INFO [train.py:901] (1/4) Epoch 12, batch 3700, loss[loss=0.2715, simple_loss=0.3348, pruned_loss=0.1041, over 7809.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3063, pruned_loss=0.07769, over 1603936.57 frames. ], batch size: 20, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 11:58:28,535 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 11:58:30,702 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92624.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:58:44,261 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92643.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:58:48,467 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:58:50,894 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.476e+02 3.116e+02 4.152e+02 8.400e+02, threshold=6.233e+02, percent-clipped=9.0 +2023-02-06 11:58:54,984 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3337, 1.9364, 3.3310, 1.6628, 2.5692, 3.7314, 3.6659, 3.2533], + device='cuda:1'), covar=tensor([0.0959, 0.1453, 0.0438, 0.2087, 0.1061, 0.0239, 0.0560, 0.0583], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0302, 0.0264, 0.0298, 0.0280, 0.0240, 0.0354, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 11:58:57,877 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7456, 2.2109, 3.5249, 2.5805, 3.0995, 2.3444, 2.0214, 1.8893], + device='cuda:1'), covar=tensor([0.4024, 0.4557, 0.1298, 0.3026, 0.2227, 0.2512, 0.1778, 0.4716], + device='cuda:1'), in_proj_covar=tensor([0.0896, 0.0874, 0.0731, 0.0854, 0.0939, 0.0809, 0.0706, 0.0770], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:58:59,710 INFO [train.py:901] (1/4) Epoch 12, batch 3750, loss[loss=0.2498, simple_loss=0.3204, pruned_loss=0.0896, over 8523.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3064, pruned_loss=0.07737, over 1608024.67 frames. ], batch size: 28, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 11:59:01,852 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1436, 1.5265, 4.2625, 1.4489, 3.7435, 3.5097, 3.8085, 3.6798], + device='cuda:1'), covar=tensor([0.0510, 0.4318, 0.0526, 0.3876, 0.1131, 0.0927, 0.0603, 0.0707], + device='cuda:1'), in_proj_covar=tensor([0.0503, 0.0579, 0.0600, 0.0540, 0.0620, 0.0533, 0.0524, 0.0585], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 11:59:17,097 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92688.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:59:28,224 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3003, 1.6399, 1.6318, 0.9589, 1.7071, 1.2410, 0.2419, 1.4769], + device='cuda:1'), covar=tensor([0.0343, 0.0218, 0.0189, 0.0349, 0.0239, 0.0679, 0.0560, 0.0169], + device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0337, 0.0286, 0.0400, 0.0328, 0.0489, 0.0365, 0.0367], + device='cuda:1'), out_proj_covar=tensor([1.1572e-04, 9.3237e-05, 7.9089e-05, 1.1163e-04, 9.1948e-05, 1.4776e-04, + 1.0352e-04, 1.0316e-04], device='cuda:1') +2023-02-06 11:59:34,453 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92713.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:59:34,943 INFO [train.py:901] (1/4) Epoch 12, batch 3800, loss[loss=0.2811, simple_loss=0.3482, pruned_loss=0.107, over 6856.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3056, pruned_loss=0.07664, over 1611208.06 frames. ], batch size: 71, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 11:59:35,189 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92714.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:59:52,022 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92738.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:59:52,766 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92739.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:59:59,547 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8788, 1.6687, 5.9865, 2.1907, 5.4444, 5.0097, 5.5726, 5.4924], + device='cuda:1'), covar=tensor([0.0418, 0.4182, 0.0335, 0.3303, 0.0859, 0.0682, 0.0407, 0.0410], + device='cuda:1'), in_proj_covar=tensor([0.0500, 0.0577, 0.0597, 0.0536, 0.0615, 0.0528, 0.0521, 0.0581], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:00:02,126 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.663e+02 2.558e+02 2.972e+02 3.756e+02 9.318e+02, threshold=5.944e+02, percent-clipped=5.0 +2023-02-06 12:00:09,490 INFO [train.py:901] (1/4) Epoch 12, batch 3850, loss[loss=0.2504, simple_loss=0.3347, pruned_loss=0.08302, over 8251.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3068, pruned_loss=0.07749, over 1608294.21 frames. ], batch size: 24, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 12:00:33,554 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 12:00:43,232 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0678, 1.7885, 3.3109, 1.3970, 2.3235, 3.6511, 3.6469, 3.0036], + device='cuda:1'), covar=tensor([0.0977, 0.1395, 0.0362, 0.2074, 0.0925, 0.0245, 0.0471, 0.0671], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0299, 0.0265, 0.0296, 0.0277, 0.0239, 0.0354, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 12:00:45,108 INFO [train.py:901] (1/4) Epoch 12, batch 3900, loss[loss=0.2224, simple_loss=0.3043, pruned_loss=0.07019, over 8457.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3071, pruned_loss=0.07748, over 1615319.48 frames. ], batch size: 25, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 12:01:08,874 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92849.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:01:11,295 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.559e+02 2.538e+02 2.989e+02 3.922e+02 7.912e+02, threshold=5.979e+02, percent-clipped=3.0 +2023-02-06 12:01:18,296 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6404, 2.4341, 4.4329, 1.3684, 2.9942, 2.2570, 1.7817, 2.7030], + device='cuda:1'), covar=tensor([0.1716, 0.2057, 0.0768, 0.3875, 0.1568, 0.2727, 0.1732, 0.2389], + device='cuda:1'), in_proj_covar=tensor([0.0492, 0.0529, 0.0539, 0.0583, 0.0622, 0.0564, 0.0477, 0.0615], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:01:19,449 INFO [train.py:901] (1/4) Epoch 12, batch 3950, loss[loss=0.2176, simple_loss=0.3116, pruned_loss=0.06174, over 8195.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3076, pruned_loss=0.07806, over 1612250.02 frames. ], batch size: 23, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:01:54,553 INFO [train.py:901] (1/4) Epoch 12, batch 4000, loss[loss=0.2105, simple_loss=0.2985, pruned_loss=0.06127, over 8299.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3072, pruned_loss=0.07789, over 1612200.50 frames. ], batch size: 23, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:01:56,841 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92917.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:02:18,344 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92949.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:02:20,893 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.373e+02 3.059e+02 3.649e+02 8.513e+02, threshold=6.118e+02, percent-clipped=6.0 +2023-02-06 12:02:28,381 INFO [train.py:901] (1/4) Epoch 12, batch 4050, loss[loss=0.2076, simple_loss=0.2879, pruned_loss=0.06369, over 7707.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3072, pruned_loss=0.07752, over 1611793.95 frames. ], batch size: 18, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:02:44,203 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92987.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:02:48,264 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:02:57,122 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1186, 2.7644, 3.4807, 2.2350, 2.0305, 3.5006, 0.6116, 2.1222], + device='cuda:1'), covar=tensor([0.2443, 0.1637, 0.0507, 0.2667, 0.3521, 0.0410, 0.4241, 0.2142], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0174, 0.0103, 0.0216, 0.0253, 0.0107, 0.0162, 0.0168], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 12:03:03,723 INFO [train.py:901] (1/4) Epoch 12, batch 4100, loss[loss=0.2522, simple_loss=0.3246, pruned_loss=0.08993, over 7127.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3074, pruned_loss=0.07756, over 1610729.79 frames. ], batch size: 72, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:03:04,562 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8056, 1.3592, 1.5140, 1.1553, 0.8434, 1.2792, 1.5375, 1.3896], + device='cuda:1'), covar=tensor([0.0534, 0.1284, 0.1766, 0.1478, 0.0645, 0.1519, 0.0743, 0.0638], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0153, 0.0194, 0.0160, 0.0104, 0.0164, 0.0117, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 12:03:13,912 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93028.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:03:21,019 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 12:03:30,614 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.418e+02 3.048e+02 3.757e+02 7.047e+02, threshold=6.097e+02, percent-clipped=3.0 +2023-02-06 12:03:31,424 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93054.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:03:37,938 INFO [train.py:901] (1/4) Epoch 12, batch 4150, loss[loss=0.2208, simple_loss=0.3114, pruned_loss=0.0651, over 8491.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3078, pruned_loss=0.07772, over 1614988.27 frames. ], batch size: 26, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:03:50,875 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:04:02,150 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 12:04:04,581 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93102.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:04:12,239 INFO [train.py:901] (1/4) Epoch 12, batch 4200, loss[loss=0.2209, simple_loss=0.2983, pruned_loss=0.07173, over 7649.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3067, pruned_loss=0.07758, over 1611235.16 frames. ], batch size: 19, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:04:24,919 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 12:04:26,410 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93133.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:04:40,114 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.730e+02 2.563e+02 2.943e+02 3.717e+02 8.503e+02, threshold=5.885e+02, percent-clipped=3.0 +2023-02-06 12:04:47,434 INFO [train.py:901] (1/4) Epoch 12, batch 4250, loss[loss=0.2689, simple_loss=0.338, pruned_loss=0.0999, over 8449.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3051, pruned_loss=0.07674, over 1607836.13 frames. ], batch size: 27, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:04:48,800 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 12:04:53,301 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 12:05:06,764 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93193.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:05:09,354 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93197.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:05:21,301 INFO [train.py:901] (1/4) Epoch 12, batch 4300, loss[loss=0.1868, simple_loss=0.275, pruned_loss=0.04929, over 8250.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3049, pruned_loss=0.07616, over 1609709.02 frames. ], batch size: 24, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:05:22,139 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6288, 1.3983, 4.7761, 1.7424, 4.3000, 3.9673, 4.3109, 4.1399], + device='cuda:1'), covar=tensor([0.0374, 0.3995, 0.0380, 0.3072, 0.0839, 0.0734, 0.0425, 0.0502], + device='cuda:1'), in_proj_covar=tensor([0.0501, 0.0580, 0.0594, 0.0537, 0.0615, 0.0527, 0.0519, 0.0581], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:05:48,566 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.718e+02 3.236e+02 4.116e+02 1.260e+03, threshold=6.473e+02, percent-clipped=7.0 +2023-02-06 12:05:54,513 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93261.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:05:56,509 INFO [train.py:901] (1/4) Epoch 12, batch 4350, loss[loss=0.2792, simple_loss=0.3509, pruned_loss=0.1038, over 8482.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3074, pruned_loss=0.07739, over 1613232.62 frames. ], batch size: 29, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:06:15,791 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93293.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:06:16,412 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 12:06:25,929 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93308.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:06:29,908 INFO [train.py:901] (1/4) Epoch 12, batch 4400, loss[loss=0.1832, simple_loss=0.2606, pruned_loss=0.0529, over 7698.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3078, pruned_loss=0.07773, over 1607097.26 frames. ], batch size: 18, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:06:46,137 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93337.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:06:58,348 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.551e+02 2.995e+02 3.715e+02 7.484e+02, threshold=5.990e+02, percent-clipped=1.0 +2023-02-06 12:06:58,373 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 12:07:01,790 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93358.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:05,648 INFO [train.py:901] (1/4) Epoch 12, batch 4450, loss[loss=0.228, simple_loss=0.3084, pruned_loss=0.07382, over 8101.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3068, pruned_loss=0.07726, over 1607193.21 frames. ], batch size: 23, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:07:11,752 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93372.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:14,528 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:19,342 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93383.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:29,363 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:36,061 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93408.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:38,776 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1938, 2.5454, 2.9450, 1.3396, 3.0164, 1.8107, 1.6156, 1.7431], + device='cuda:1'), covar=tensor([0.0590, 0.0308, 0.0248, 0.0605, 0.0368, 0.0657, 0.0689, 0.0482], + device='cuda:1'), in_proj_covar=tensor([0.0392, 0.0328, 0.0273, 0.0387, 0.0320, 0.0474, 0.0356, 0.0357], + device='cuda:1'), out_proj_covar=tensor([1.1110e-04, 9.0557e-05, 7.5531e-05, 1.0777e-04, 8.9835e-05, 1.4287e-04, + 1.0068e-04, 1.0019e-04], device='cuda:1') +2023-02-06 12:07:39,490 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5692, 2.0626, 3.0417, 2.3693, 2.9632, 2.2809, 1.9470, 1.3950], + device='cuda:1'), covar=tensor([0.3560, 0.3749, 0.1306, 0.2771, 0.1654, 0.2113, 0.1584, 0.4278], + device='cuda:1'), in_proj_covar=tensor([0.0892, 0.0871, 0.0730, 0.0852, 0.0931, 0.0803, 0.0705, 0.0768], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:07:39,925 INFO [train.py:901] (1/4) Epoch 12, batch 4500, loss[loss=0.1626, simple_loss=0.2445, pruned_loss=0.0403, over 7416.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3063, pruned_loss=0.07754, over 1604257.92 frames. ], batch size: 17, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:07:44,325 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.89 vs. limit=5.0 +2023-02-06 12:07:50,737 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 12:08:06,003 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93452.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:06,475 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.576e+02 3.193e+02 4.187e+02 6.619e+02, threshold=6.386e+02, percent-clipped=4.0 +2023-02-06 12:08:06,709 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93453.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:13,866 INFO [train.py:901] (1/4) Epoch 12, batch 4550, loss[loss=0.2418, simple_loss=0.3169, pruned_loss=0.08333, over 7925.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3059, pruned_loss=0.07734, over 1605913.16 frames. ], batch size: 20, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:08:24,178 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:25,008 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:31,261 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6539, 1.9126, 2.0810, 1.1713, 2.1512, 1.4112, 0.5583, 1.7478], + device='cuda:1'), covar=tensor([0.0355, 0.0208, 0.0165, 0.0329, 0.0272, 0.0599, 0.0564, 0.0191], + device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0331, 0.0276, 0.0390, 0.0322, 0.0476, 0.0359, 0.0359], + device='cuda:1'), out_proj_covar=tensor([1.1197e-04, 9.1400e-05, 7.6214e-05, 1.0852e-04, 9.0395e-05, 1.4353e-04, + 1.0154e-04, 1.0079e-04], device='cuda:1') +2023-02-06 12:08:31,907 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:49,683 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93513.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:50,198 INFO [train.py:901] (1/4) Epoch 12, batch 4600, loss[loss=0.2076, simple_loss=0.2823, pruned_loss=0.06643, over 7799.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3063, pruned_loss=0.0775, over 1604666.85 frames. ], batch size: 19, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:08:53,906 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 12:09:16,615 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.455e+02 3.020e+02 4.052e+02 9.299e+02, threshold=6.041e+02, percent-clipped=5.0 +2023-02-06 12:09:18,236 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7034, 1.4468, 1.5192, 1.2217, 0.8646, 1.2609, 1.4910, 1.2500], + device='cuda:1'), covar=tensor([0.0545, 0.1262, 0.1774, 0.1448, 0.0621, 0.1618, 0.0711, 0.0687], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0192, 0.0158, 0.0104, 0.0163, 0.0116, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 12:09:24,877 INFO [train.py:901] (1/4) Epoch 12, batch 4650, loss[loss=0.2833, simple_loss=0.3338, pruned_loss=0.1164, over 7233.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3065, pruned_loss=0.0774, over 1609594.43 frames. ], batch size: 16, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:09:25,092 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93564.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:09:41,890 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93589.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:09:45,159 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:09:59,883 INFO [train.py:901] (1/4) Epoch 12, batch 4700, loss[loss=0.2407, simple_loss=0.3111, pruned_loss=0.08514, over 8597.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3062, pruned_loss=0.07706, over 1610489.14 frames. ], batch size: 31, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:10:12,704 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93632.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:10:21,477 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3183, 1.8498, 2.7212, 2.1671, 2.4930, 2.1580, 1.7097, 1.1736], + device='cuda:1'), covar=tensor([0.3909, 0.3956, 0.1064, 0.2474, 0.1724, 0.2069, 0.1797, 0.3786], + device='cuda:1'), in_proj_covar=tensor([0.0893, 0.0871, 0.0725, 0.0851, 0.0934, 0.0802, 0.0699, 0.0767], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:10:26,597 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.440e+02 2.370e+02 2.939e+02 3.568e+02 8.447e+02, threshold=5.879e+02, percent-clipped=4.0 +2023-02-06 12:10:29,424 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93657.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:10:33,905 INFO [train.py:901] (1/4) Epoch 12, batch 4750, loss[loss=0.2155, simple_loss=0.2892, pruned_loss=0.07093, over 7818.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3062, pruned_loss=0.07774, over 1611067.71 frames. ], batch size: 20, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:10:34,142 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93664.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:10:51,742 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93689.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:00,453 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 12:11:02,146 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 +2023-02-06 12:11:02,482 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 12:11:04,757 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93708.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:09,466 INFO [train.py:901] (1/4) Epoch 12, batch 4800, loss[loss=0.2276, simple_loss=0.3094, pruned_loss=0.07291, over 8088.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3052, pruned_loss=0.07714, over 1610960.40 frames. ], batch size: 21, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:11:23,305 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:30,151 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93743.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:36,833 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.527e+02 2.519e+02 2.967e+02 3.635e+02 7.460e+02, threshold=5.934e+02, percent-clipped=2.0 +2023-02-06 12:11:44,128 INFO [train.py:901] (1/4) Epoch 12, batch 4850, loss[loss=0.2949, simple_loss=0.3635, pruned_loss=0.1131, over 8103.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3053, pruned_loss=0.07717, over 1610601.19 frames. ], batch size: 23, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:11:44,993 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7931, 1.7242, 2.6537, 1.3943, 2.1547, 2.9213, 2.9305, 2.4916], + device='cuda:1'), covar=tensor([0.0930, 0.1214, 0.0533, 0.1898, 0.1056, 0.0306, 0.0692, 0.0606], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0299, 0.0265, 0.0295, 0.0279, 0.0240, 0.0354, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 12:11:47,093 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93768.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:47,777 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93769.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:53,107 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 12:12:00,552 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93788.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:12:04,721 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:12:18,633 INFO [train.py:901] (1/4) Epoch 12, batch 4900, loss[loss=0.2262, simple_loss=0.2954, pruned_loss=0.07846, over 7973.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3057, pruned_loss=0.07698, over 1611368.03 frames. ], batch size: 21, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:12:42,679 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:12:45,718 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.559e+02 2.389e+02 2.920e+02 3.679e+02 7.315e+02, threshold=5.841e+02, percent-clipped=3.0 +2023-02-06 12:12:53,902 INFO [train.py:901] (1/4) Epoch 12, batch 4950, loss[loss=0.2138, simple_loss=0.2918, pruned_loss=0.06796, over 8276.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3064, pruned_loss=0.07745, over 1610721.56 frames. ], batch size: 23, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:13:00,015 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93873.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:13:27,063 INFO [train.py:901] (1/4) Epoch 12, batch 5000, loss[loss=0.2406, simple_loss=0.3142, pruned_loss=0.08352, over 8339.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3073, pruned_loss=0.07765, over 1610825.03 frames. ], batch size: 26, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:13:55,386 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.574e+02 3.082e+02 3.748e+02 7.333e+02, threshold=6.165e+02, percent-clipped=4.0 +2023-02-06 12:14:02,944 INFO [train.py:901] (1/4) Epoch 12, batch 5050, loss[loss=0.294, simple_loss=0.3475, pruned_loss=0.1203, over 8089.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3074, pruned_loss=0.07806, over 1611000.07 frames. ], batch size: 21, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:14:27,666 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:14:31,162 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 12:14:38,585 INFO [train.py:901] (1/4) Epoch 12, batch 5100, loss[loss=0.219, simple_loss=0.2932, pruned_loss=0.07244, over 7980.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3075, pruned_loss=0.0782, over 1611882.58 frames. ], batch size: 21, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:15:05,336 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.576e+02 2.962e+02 4.029e+02 5.912e+02, threshold=5.924e+02, percent-clipped=0.0 +2023-02-06 12:15:13,500 INFO [train.py:901] (1/4) Epoch 12, batch 5150, loss[loss=0.2218, simple_loss=0.2832, pruned_loss=0.08018, over 7515.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3064, pruned_loss=0.07788, over 1611930.79 frames. ], batch size: 18, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:15:33,466 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 12:15:46,359 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9109, 2.2940, 3.6455, 1.6212, 2.9750, 2.3293, 2.0680, 2.6842], + device='cuda:1'), covar=tensor([0.1509, 0.2207, 0.0649, 0.3782, 0.1271, 0.2463, 0.1709, 0.2039], + device='cuda:1'), in_proj_covar=tensor([0.0492, 0.0526, 0.0535, 0.0587, 0.0618, 0.0558, 0.0478, 0.0617], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:15:47,487 INFO [train.py:901] (1/4) Epoch 12, batch 5200, loss[loss=0.206, simple_loss=0.2739, pruned_loss=0.06905, over 7217.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3073, pruned_loss=0.07802, over 1611794.34 frames. ], batch size: 16, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:15:58,685 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3513, 1.4588, 1.2994, 1.8788, 0.7750, 1.1874, 1.2320, 1.4711], + device='cuda:1'), covar=tensor([0.0926, 0.0909, 0.1165, 0.0486, 0.1216, 0.1586, 0.0917, 0.0796], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0216, 0.0258, 0.0219, 0.0220, 0.0256, 0.0263, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 12:15:59,925 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94132.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:16:00,737 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8296, 1.7354, 2.9029, 1.3498, 2.2181, 3.0909, 3.1445, 2.6564], + device='cuda:1'), covar=tensor([0.0871, 0.1214, 0.0363, 0.1801, 0.0781, 0.0278, 0.0542, 0.0614], + device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0296, 0.0262, 0.0291, 0.0275, 0.0238, 0.0349, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 12:16:14,668 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.571e+02 3.074e+02 4.467e+02 8.286e+02, threshold=6.149e+02, percent-clipped=7.0 +2023-02-06 12:16:21,922 INFO [train.py:901] (1/4) Epoch 12, batch 5250, loss[loss=0.1992, simple_loss=0.2643, pruned_loss=0.06704, over 7552.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3075, pruned_loss=0.07812, over 1612385.43 frames. ], batch size: 18, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:16:25,911 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 12:16:57,674 INFO [train.py:901] (1/4) Epoch 12, batch 5300, loss[loss=0.223, simple_loss=0.3047, pruned_loss=0.07061, over 8043.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3064, pruned_loss=0.07694, over 1611523.04 frames. ], batch size: 22, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:17:13,241 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 12:17:15,484 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94241.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:17:19,507 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:17:23,399 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.628e+02 3.237e+02 4.138e+02 9.258e+02, threshold=6.473e+02, percent-clipped=5.0 +2023-02-06 12:17:31,605 INFO [train.py:901] (1/4) Epoch 12, batch 5350, loss[loss=0.2617, simple_loss=0.3355, pruned_loss=0.0939, over 8497.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3058, pruned_loss=0.07639, over 1613789.11 frames. ], batch size: 29, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:17:34,488 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3204, 1.9528, 2.7410, 2.3071, 2.5889, 2.0982, 1.9457, 1.8484], + device='cuda:1'), covar=tensor([0.3339, 0.3585, 0.1206, 0.2208, 0.1530, 0.2274, 0.1578, 0.3127], + device='cuda:1'), in_proj_covar=tensor([0.0886, 0.0863, 0.0722, 0.0844, 0.0929, 0.0795, 0.0697, 0.0758], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:18:05,212 INFO [train.py:901] (1/4) Epoch 12, batch 5400, loss[loss=0.2261, simple_loss=0.3086, pruned_loss=0.07179, over 8467.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3062, pruned_loss=0.07657, over 1613718.67 frames. ], batch size: 25, lr: 6.33e-03, grad_scale: 8.0 +2023-02-06 12:18:23,610 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2536, 3.0953, 2.8471, 1.6084, 2.8850, 2.8009, 2.8121, 2.6981], + device='cuda:1'), covar=tensor([0.1328, 0.1051, 0.1622, 0.5355, 0.1308, 0.1448, 0.1955, 0.1220], + device='cuda:1'), in_proj_covar=tensor([0.0465, 0.0379, 0.0389, 0.0484, 0.0379, 0.0386, 0.0381, 0.0333], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:18:25,506 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:18:32,244 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.485e+02 2.978e+02 4.110e+02 9.009e+02, threshold=5.957e+02, percent-clipped=6.0 +2023-02-06 12:18:34,961 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 12:18:39,987 INFO [train.py:901] (1/4) Epoch 12, batch 5450, loss[loss=0.2207, simple_loss=0.2985, pruned_loss=0.07148, over 5127.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3061, pruned_loss=0.07684, over 1609936.06 frames. ], batch size: 11, lr: 6.33e-03, grad_scale: 8.0 +2023-02-06 12:18:41,434 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:18:47,042 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.14 vs. limit=5.0 +2023-02-06 12:18:57,715 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.3671, 1.2096, 3.6585, 1.4185, 2.7635, 2.7770, 3.2509, 3.2337], + device='cuda:1'), covar=tensor([0.1648, 0.7093, 0.1531, 0.5369, 0.3178, 0.2227, 0.1553, 0.1487], + device='cuda:1'), in_proj_covar=tensor([0.0505, 0.0580, 0.0593, 0.0537, 0.0619, 0.0529, 0.0516, 0.0578], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:19:12,399 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 12:19:15,765 INFO [train.py:901] (1/4) Epoch 12, batch 5500, loss[loss=0.2537, simple_loss=0.3315, pruned_loss=0.0879, over 8327.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3061, pruned_loss=0.07648, over 1609412.68 frames. ], batch size: 26, lr: 6.33e-03, grad_scale: 16.0 +2023-02-06 12:19:30,249 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.03 vs. limit=5.0 +2023-02-06 12:19:43,164 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.406e+02 2.798e+02 3.361e+02 6.650e+02, threshold=5.597e+02, percent-clipped=1.0 +2023-02-06 12:19:45,380 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94458.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:19:49,115 INFO [train.py:901] (1/4) Epoch 12, batch 5550, loss[loss=0.2806, simple_loss=0.3371, pruned_loss=0.112, over 8541.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3064, pruned_loss=0.07709, over 1609694.40 frames. ], batch size: 28, lr: 6.33e-03, grad_scale: 4.0 +2023-02-06 12:20:16,694 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94503.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:20:24,052 INFO [train.py:901] (1/4) Epoch 12, batch 5600, loss[loss=0.1878, simple_loss=0.2707, pruned_loss=0.05246, over 7816.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3056, pruned_loss=0.07684, over 1611072.75 frames. ], batch size: 20, lr: 6.33e-03, grad_scale: 8.0 +2023-02-06 12:20:35,266 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94528.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:20:54,480 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.675e+02 3.313e+02 4.214e+02 1.006e+03, threshold=6.626e+02, percent-clipped=7.0 +2023-02-06 12:21:00,674 INFO [train.py:901] (1/4) Epoch 12, batch 5650, loss[loss=0.218, simple_loss=0.3051, pruned_loss=0.06547, over 8032.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.305, pruned_loss=0.07612, over 1610713.76 frames. ], batch size: 22, lr: 6.33e-03, grad_scale: 8.0 +2023-02-06 12:21:15,257 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94585.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:21:21,336 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 12:21:35,767 INFO [train.py:901] (1/4) Epoch 12, batch 5700, loss[loss=0.2085, simple_loss=0.2894, pruned_loss=0.06382, over 7968.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3048, pruned_loss=0.076, over 1607172.29 frames. ], batch size: 21, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:22:04,739 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.327e+02 3.036e+02 3.801e+02 7.493e+02, threshold=6.072e+02, percent-clipped=2.0 +2023-02-06 12:22:10,812 INFO [train.py:901] (1/4) Epoch 12, batch 5750, loss[loss=0.2196, simple_loss=0.3069, pruned_loss=0.06612, over 8365.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3038, pruned_loss=0.07576, over 1609035.33 frames. ], batch size: 24, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:22:15,197 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8672, 1.7734, 2.3059, 1.6956, 1.2239, 2.3782, 0.4840, 1.3611], + device='cuda:1'), covar=tensor([0.2317, 0.1727, 0.0496, 0.1965, 0.4035, 0.0428, 0.3263, 0.1877], + device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0169, 0.0100, 0.0213, 0.0252, 0.0104, 0.0163, 0.0164], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 12:22:26,195 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 12:22:35,724 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94700.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:22:42,477 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94710.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:22:45,141 INFO [train.py:901] (1/4) Epoch 12, batch 5800, loss[loss=0.258, simple_loss=0.3485, pruned_loss=0.08374, over 8761.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3043, pruned_loss=0.07573, over 1609495.62 frames. ], batch size: 30, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:22:45,332 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94714.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:23:02,548 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94739.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:23:13,645 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.575e+02 3.288e+02 4.021e+02 7.847e+02, threshold=6.576e+02, percent-clipped=2.0 +2023-02-06 12:23:19,972 INFO [train.py:901] (1/4) Epoch 12, batch 5850, loss[loss=0.3504, simple_loss=0.3878, pruned_loss=0.1565, over 6802.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3043, pruned_loss=0.07593, over 1608586.91 frames. ], batch size: 72, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:23:26,199 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3800, 1.6609, 1.7625, 0.9561, 1.7023, 1.4066, 0.2618, 1.5991], + device='cuda:1'), covar=tensor([0.0314, 0.0223, 0.0184, 0.0321, 0.0289, 0.0618, 0.0581, 0.0164], + device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0328, 0.0275, 0.0386, 0.0320, 0.0475, 0.0356, 0.0357], + device='cuda:1'), out_proj_covar=tensor([1.1132e-04, 9.0097e-05, 7.6090e-05, 1.0710e-04, 8.9457e-05, 1.4311e-04, + 1.0049e-04, 1.0016e-04], device='cuda:1') +2023-02-06 12:23:54,266 INFO [train.py:901] (1/4) Epoch 12, batch 5900, loss[loss=0.2229, simple_loss=0.2824, pruned_loss=0.08163, over 7796.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3046, pruned_loss=0.07598, over 1607149.80 frames. ], batch size: 19, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:24:01,747 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94825.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:24:22,270 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.554e+02 2.604e+02 3.248e+02 4.213e+02 6.479e+02, threshold=6.496e+02, percent-clipped=0.0 +2023-02-06 12:24:28,372 INFO [train.py:901] (1/4) Epoch 12, batch 5950, loss[loss=0.2365, simple_loss=0.3279, pruned_loss=0.07253, over 8355.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3049, pruned_loss=0.07629, over 1610350.50 frames. ], batch size: 24, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:25:03,798 INFO [train.py:901] (1/4) Epoch 12, batch 6000, loss[loss=0.2492, simple_loss=0.3278, pruned_loss=0.08525, over 8563.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3051, pruned_loss=0.07637, over 1606121.02 frames. ], batch size: 34, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:25:03,799 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 12:25:16,952 INFO [train.py:935] (1/4) Epoch 12, validation: loss=0.1862, simple_loss=0.286, pruned_loss=0.04318, over 944034.00 frames. +2023-02-06 12:25:16,953 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 12:25:17,910 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2788, 2.5684, 3.1026, 1.4450, 3.2273, 1.8505, 1.4250, 2.0926], + device='cuda:1'), covar=tensor([0.0600, 0.0304, 0.0164, 0.0540, 0.0292, 0.0666, 0.0704, 0.0402], + device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0327, 0.0276, 0.0386, 0.0320, 0.0474, 0.0357, 0.0356], + device='cuda:1'), out_proj_covar=tensor([1.1110e-04, 9.0045e-05, 7.6154e-05, 1.0725e-04, 8.9410e-05, 1.4274e-04, + 1.0059e-04, 9.9550e-05], device='cuda:1') +2023-02-06 12:25:44,732 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.831e+02 2.442e+02 2.970e+02 3.787e+02 9.017e+02, threshold=5.940e+02, percent-clipped=3.0 +2023-02-06 12:25:45,512 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94956.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:25:50,745 INFO [train.py:901] (1/4) Epoch 12, batch 6050, loss[loss=0.2375, simple_loss=0.3142, pruned_loss=0.08037, over 8081.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3046, pruned_loss=0.07585, over 1603879.23 frames. ], batch size: 21, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:26:02,527 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94981.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:26:03,821 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9596, 1.7365, 1.8315, 1.5815, 1.0359, 1.5952, 1.9485, 2.0644], + device='cuda:1'), covar=tensor([0.0454, 0.1173, 0.1679, 0.1323, 0.0629, 0.1456, 0.0674, 0.0549], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0192, 0.0158, 0.0103, 0.0162, 0.0115, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 12:26:25,547 INFO [train.py:901] (1/4) Epoch 12, batch 6100, loss[loss=0.2043, simple_loss=0.2874, pruned_loss=0.06059, over 7915.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3055, pruned_loss=0.07641, over 1604279.60 frames. ], batch size: 20, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:26:38,777 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7300, 2.0158, 2.2654, 1.3510, 2.3537, 1.4323, 0.6659, 1.9839], + device='cuda:1'), covar=tensor([0.0442, 0.0234, 0.0171, 0.0390, 0.0306, 0.0688, 0.0623, 0.0202], + device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0330, 0.0280, 0.0390, 0.0324, 0.0479, 0.0360, 0.0359], + device='cuda:1'), out_proj_covar=tensor([1.1226e-04, 9.0735e-05, 7.7408e-05, 1.0852e-04, 9.0699e-05, 1.4404e-04, + 1.0164e-04, 1.0065e-04], device='cuda:1') +2023-02-06 12:26:54,026 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 12:26:54,678 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.590e+02 3.216e+02 4.301e+02 8.648e+02, threshold=6.433e+02, percent-clipped=2.0 +2023-02-06 12:27:00,775 INFO [train.py:901] (1/4) Epoch 12, batch 6150, loss[loss=0.2556, simple_loss=0.3329, pruned_loss=0.08919, over 8188.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3086, pruned_loss=0.07822, over 1612494.99 frames. ], batch size: 23, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:27:12,216 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95081.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:27:29,629 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95106.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:27:34,674 INFO [train.py:901] (1/4) Epoch 12, batch 6200, loss[loss=0.2633, simple_loss=0.3385, pruned_loss=0.09409, over 8108.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.309, pruned_loss=0.07903, over 1614834.89 frames. ], batch size: 23, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:27:41,092 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95123.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:28:04,335 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 2.519e+02 2.980e+02 3.798e+02 7.393e+02, threshold=5.961e+02, percent-clipped=2.0 +2023-02-06 12:28:10,299 INFO [train.py:901] (1/4) Epoch 12, batch 6250, loss[loss=0.2226, simple_loss=0.2965, pruned_loss=0.0744, over 8218.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3088, pruned_loss=0.07872, over 1614255.17 frames. ], batch size: 49, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:28:43,837 INFO [train.py:901] (1/4) Epoch 12, batch 6300, loss[loss=0.251, simple_loss=0.327, pruned_loss=0.08752, over 8362.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3092, pruned_loss=0.07928, over 1613025.97 frames. ], batch size: 24, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:28:46,417 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.84 vs. limit=5.0 +2023-02-06 12:29:13,416 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.657e+02 3.224e+02 4.358e+02 1.571e+03, threshold=6.448e+02, percent-clipped=5.0 +2023-02-06 12:29:20,995 INFO [train.py:901] (1/4) Epoch 12, batch 6350, loss[loss=0.2465, simple_loss=0.3189, pruned_loss=0.08707, over 8496.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3081, pruned_loss=0.0794, over 1609878.48 frames. ], batch size: 29, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:29:30,734 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:29:36,286 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:29:50,041 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7364, 2.1021, 4.1722, 1.4695, 2.9501, 2.2081, 1.7442, 2.7151], + device='cuda:1'), covar=tensor([0.1760, 0.2603, 0.0606, 0.4034, 0.1695, 0.2847, 0.1948, 0.2348], + device='cuda:1'), in_proj_covar=tensor([0.0493, 0.0531, 0.0537, 0.0586, 0.0621, 0.0559, 0.0478, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:29:55,408 INFO [train.py:901] (1/4) Epoch 12, batch 6400, loss[loss=0.2275, simple_loss=0.3131, pruned_loss=0.07093, over 8284.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3077, pruned_loss=0.07827, over 1613042.01 frames. ], batch size: 23, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:30:23,572 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.403e+02 2.937e+02 3.904e+02 6.682e+02, threshold=5.874e+02, percent-clipped=3.0 +2023-02-06 12:30:25,102 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95357.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:30:27,149 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3968, 1.5744, 2.2621, 1.3067, 1.5586, 1.6848, 1.5252, 1.3449], + device='cuda:1'), covar=tensor([0.1665, 0.2068, 0.0738, 0.3704, 0.1567, 0.2687, 0.1757, 0.1839], + device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0526, 0.0533, 0.0579, 0.0617, 0.0555, 0.0472, 0.0610], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:30:29,661 INFO [train.py:901] (1/4) Epoch 12, batch 6450, loss[loss=0.2287, simple_loss=0.2876, pruned_loss=0.08494, over 7780.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3074, pruned_loss=0.07853, over 1612345.60 frames. ], batch size: 19, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:31:05,020 INFO [train.py:901] (1/4) Epoch 12, batch 6500, loss[loss=0.272, simple_loss=0.3453, pruned_loss=0.09931, over 8031.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3059, pruned_loss=0.07755, over 1611284.39 frames. ], batch size: 22, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:31:14,614 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4448, 2.0125, 3.4436, 1.3096, 2.4117, 1.8424, 1.6526, 2.3586], + device='cuda:1'), covar=tensor([0.1745, 0.2198, 0.0665, 0.3888, 0.1690, 0.2950, 0.1873, 0.2174], + device='cuda:1'), in_proj_covar=tensor([0.0489, 0.0527, 0.0533, 0.0580, 0.0618, 0.0555, 0.0473, 0.0611], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:31:31,886 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.431e+02 2.857e+02 3.846e+02 1.801e+03, threshold=5.713e+02, percent-clipped=8.0 +2023-02-06 12:31:35,001 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-02-06 12:31:37,952 INFO [train.py:901] (1/4) Epoch 12, batch 6550, loss[loss=0.1936, simple_loss=0.2762, pruned_loss=0.05543, over 7658.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.307, pruned_loss=0.07812, over 1613711.77 frames. ], batch size: 19, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:31:40,440 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.34 vs. limit=5.0 +2023-02-06 12:31:40,699 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95467.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:31:59,679 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6384, 2.3475, 4.8304, 2.8307, 4.3714, 4.1827, 4.5376, 4.4463], + device='cuda:1'), covar=tensor([0.0571, 0.3345, 0.0454, 0.2729, 0.0857, 0.0716, 0.0457, 0.0448], + device='cuda:1'), in_proj_covar=tensor([0.0512, 0.0581, 0.0594, 0.0544, 0.0622, 0.0532, 0.0520, 0.0582], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:32:06,855 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 12:32:13,527 INFO [train.py:901] (1/4) Epoch 12, batch 6600, loss[loss=0.2069, simple_loss=0.2838, pruned_loss=0.065, over 7656.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3059, pruned_loss=0.07763, over 1612703.24 frames. ], batch size: 19, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:32:25,471 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-06 12:32:25,784 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 12:32:40,269 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.522e+02 3.078e+02 3.913e+02 8.021e+02, threshold=6.157e+02, percent-clipped=7.0 +2023-02-06 12:32:43,793 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6853, 1.7516, 2.3350, 1.3791, 1.1106, 2.3795, 0.3341, 1.3155], + device='cuda:1'), covar=tensor([0.2523, 0.1461, 0.0416, 0.2731, 0.3925, 0.0349, 0.3062, 0.2067], + device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0173, 0.0103, 0.0216, 0.0256, 0.0108, 0.0163, 0.0166], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 12:32:46,211 INFO [train.py:901] (1/4) Epoch 12, batch 6650, loss[loss=0.2122, simple_loss=0.2972, pruned_loss=0.06362, over 8242.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3063, pruned_loss=0.0775, over 1615751.16 frames. ], batch size: 24, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:32:59,154 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95582.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:33:15,569 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 +2023-02-06 12:33:21,233 INFO [train.py:901] (1/4) Epoch 12, batch 6700, loss[loss=0.2353, simple_loss=0.3062, pruned_loss=0.08215, over 7438.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3059, pruned_loss=0.07698, over 1612526.21 frames. ], batch size: 17, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:33:27,483 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:33:33,688 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:33:35,025 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7255, 1.4826, 2.7795, 1.1576, 2.1028, 3.0276, 3.1211, 2.5903], + device='cuda:1'), covar=tensor([0.1075, 0.1450, 0.0427, 0.2206, 0.0881, 0.0303, 0.0569, 0.0637], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0297, 0.0261, 0.0291, 0.0272, 0.0236, 0.0347, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 12:33:37,106 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:33:50,484 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.832e+02 2.656e+02 3.142e+02 4.011e+02 7.522e+02, threshold=6.284e+02, percent-clipped=4.0 +2023-02-06 12:33:56,566 INFO [train.py:901] (1/4) Epoch 12, batch 6750, loss[loss=0.2241, simple_loss=0.3033, pruned_loss=0.0725, over 8190.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3068, pruned_loss=0.07728, over 1613538.26 frames. ], batch size: 23, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:33:59,366 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6635, 1.6179, 2.0223, 1.6013, 1.1630, 2.0429, 0.3097, 1.1920], + device='cuda:1'), covar=tensor([0.1978, 0.1407, 0.0447, 0.1407, 0.3267, 0.0402, 0.3113, 0.1641], + device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0171, 0.0102, 0.0213, 0.0253, 0.0106, 0.0161, 0.0164], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 12:34:06,791 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4866, 1.8467, 1.9044, 1.0925, 1.9643, 1.3130, 0.4246, 1.6425], + device='cuda:1'), covar=tensor([0.0347, 0.0187, 0.0182, 0.0351, 0.0230, 0.0654, 0.0557, 0.0181], + device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0329, 0.0281, 0.0391, 0.0325, 0.0482, 0.0360, 0.0358], + device='cuda:1'), out_proj_covar=tensor([1.1175e-04, 9.0035e-05, 7.7683e-05, 1.0853e-04, 9.0945e-05, 1.4494e-04, + 1.0150e-04, 1.0027e-04], device='cuda:1') +2023-02-06 12:34:22,169 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95701.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:34:30,621 INFO [train.py:901] (1/4) Epoch 12, batch 6800, loss[loss=0.212, simple_loss=0.2773, pruned_loss=0.07337, over 7185.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3068, pruned_loss=0.0771, over 1616663.96 frames. ], batch size: 16, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:34:40,704 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 12:34:44,337 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5746, 1.3967, 2.7736, 1.2607, 2.1608, 3.0295, 3.1095, 2.5755], + device='cuda:1'), covar=tensor([0.1176, 0.1664, 0.0425, 0.2241, 0.0885, 0.0316, 0.0640, 0.0703], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0298, 0.0261, 0.0292, 0.0273, 0.0236, 0.0349, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 12:34:47,117 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:34:53,226 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:35:00,265 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.510e+02 2.822e+02 3.564e+02 9.162e+02, threshold=5.644e+02, percent-clipped=3.0 +2023-02-06 12:35:06,297 INFO [train.py:901] (1/4) Epoch 12, batch 6850, loss[loss=0.2534, simple_loss=0.318, pruned_loss=0.09443, over 8476.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3058, pruned_loss=0.07598, over 1617714.70 frames. ], batch size: 27, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:35:21,827 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1325, 1.4794, 4.1507, 1.6869, 2.4912, 4.7213, 4.6809, 4.0275], + device='cuda:1'), covar=tensor([0.1190, 0.1862, 0.0314, 0.2179, 0.1127, 0.0199, 0.0443, 0.0576], + device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0297, 0.0259, 0.0291, 0.0271, 0.0235, 0.0347, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 12:35:26,873 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 12:35:40,360 INFO [train.py:901] (1/4) Epoch 12, batch 6900, loss[loss=0.2121, simple_loss=0.2769, pruned_loss=0.07368, over 7814.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3067, pruned_loss=0.07655, over 1620473.41 frames. ], batch size: 20, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:35:41,925 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95816.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:35:51,213 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95830.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:35:57,297 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95838.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:36:05,035 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2050, 3.0990, 2.8610, 1.4085, 2.8056, 2.9519, 2.9091, 2.6140], + device='cuda:1'), covar=tensor([0.1124, 0.0901, 0.1353, 0.4863, 0.1086, 0.1135, 0.1496, 0.1122], + device='cuda:1'), in_proj_covar=tensor([0.0472, 0.0384, 0.0394, 0.0488, 0.0388, 0.0389, 0.0387, 0.0334], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:36:05,634 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9413, 1.5886, 1.6636, 1.4623, 1.1489, 1.5806, 1.7544, 1.5743], + device='cuda:1'), covar=tensor([0.0510, 0.1161, 0.1591, 0.1340, 0.0622, 0.1399, 0.0710, 0.0583], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0151, 0.0190, 0.0158, 0.0103, 0.0162, 0.0114, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 12:36:08,023 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.682e+02 2.693e+02 3.422e+02 4.342e+02 1.062e+03, threshold=6.843e+02, percent-clipped=12.0 +2023-02-06 12:36:14,241 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95863.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:36:14,702 INFO [train.py:901] (1/4) Epoch 12, batch 6950, loss[loss=0.2294, simple_loss=0.3005, pruned_loss=0.07912, over 7941.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3075, pruned_loss=0.07702, over 1620836.37 frames. ], batch size: 20, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:36:15,460 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95865.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:36:20,070 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8471, 1.4741, 3.3852, 1.3444, 2.3707, 3.7771, 3.8862, 3.0648], + device='cuda:1'), covar=tensor([0.1153, 0.1730, 0.0394, 0.2275, 0.0984, 0.0286, 0.0440, 0.0735], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0298, 0.0261, 0.0292, 0.0273, 0.0237, 0.0350, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 12:36:34,624 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 12:36:48,738 INFO [train.py:901] (1/4) Epoch 12, batch 7000, loss[loss=0.2209, simple_loss=0.2994, pruned_loss=0.0712, over 8506.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3081, pruned_loss=0.07774, over 1620147.21 frames. ], batch size: 39, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:36:58,642 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 12:37:17,431 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.501e+02 3.116e+02 3.850e+02 8.001e+02, threshold=6.232e+02, percent-clipped=2.0 +2023-02-06 12:37:22,085 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6325, 4.5992, 4.2075, 1.9388, 4.1141, 4.1777, 4.1488, 3.8672], + device='cuda:1'), covar=tensor([0.0786, 0.0610, 0.1146, 0.4667, 0.0890, 0.0742, 0.1374, 0.0603], + device='cuda:1'), in_proj_covar=tensor([0.0473, 0.0383, 0.0394, 0.0486, 0.0386, 0.0388, 0.0386, 0.0335], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:37:23,288 INFO [train.py:901] (1/4) Epoch 12, batch 7050, loss[loss=0.2444, simple_loss=0.3138, pruned_loss=0.08748, over 7817.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3076, pruned_loss=0.07791, over 1619323.43 frames. ], batch size: 20, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:37:34,014 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:37:44,071 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:37:47,395 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95998.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:37:50,531 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96001.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:37:58,505 INFO [train.py:901] (1/4) Epoch 12, batch 7100, loss[loss=0.2806, simple_loss=0.346, pruned_loss=0.1076, over 8106.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3088, pruned_loss=0.07881, over 1620574.58 frames. ], batch size: 23, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:38:01,398 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:06,474 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.79 vs. limit=5.0 +2023-02-06 12:38:07,001 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96026.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:20,456 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8817, 1.8517, 2.4000, 1.6584, 1.2945, 2.4260, 0.4119, 1.4527], + device='cuda:1'), covar=tensor([0.2674, 0.1671, 0.0415, 0.1874, 0.4103, 0.0461, 0.3503, 0.1950], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0176, 0.0104, 0.0219, 0.0257, 0.0110, 0.0166, 0.0168], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 12:38:26,975 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.539e+02 3.029e+02 4.080e+02 8.783e+02, threshold=6.058e+02, percent-clipped=4.0 +2023-02-06 12:38:33,144 INFO [train.py:901] (1/4) Epoch 12, batch 7150, loss[loss=0.219, simple_loss=0.287, pruned_loss=0.07551, over 7540.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3088, pruned_loss=0.07872, over 1619343.62 frames. ], batch size: 18, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:38:38,700 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96072.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:42,728 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8923, 2.5133, 3.3165, 1.6782, 1.5374, 3.3358, 0.5072, 1.8290], + device='cuda:1'), covar=tensor([0.1996, 0.1392, 0.0450, 0.3006, 0.4048, 0.0355, 0.3903, 0.2114], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0176, 0.0104, 0.0219, 0.0257, 0.0110, 0.0166, 0.0167], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 12:38:54,774 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:56,848 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96097.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:58,164 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96098.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:39:08,659 INFO [train.py:901] (1/4) Epoch 12, batch 7200, loss[loss=0.2525, simple_loss=0.329, pruned_loss=0.088, over 8296.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3081, pruned_loss=0.07881, over 1614996.98 frames. ], batch size: 23, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:39:36,231 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.446e+02 3.002e+02 3.633e+02 6.248e+02, threshold=6.005e+02, percent-clipped=1.0 +2023-02-06 12:39:42,865 INFO [train.py:901] (1/4) Epoch 12, batch 7250, loss[loss=0.296, simple_loss=0.3692, pruned_loss=0.1114, over 8435.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3079, pruned_loss=0.07834, over 1617782.63 frames. ], batch size: 27, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:39:49,510 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96174.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:39:55,426 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96183.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:40:14,133 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96209.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:40:14,950 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5976, 2.2484, 3.3598, 2.6777, 3.0445, 2.3543, 2.0473, 1.9581], + device='cuda:1'), covar=tensor([0.3589, 0.3960, 0.1286, 0.2588, 0.1928, 0.2061, 0.1593, 0.3912], + device='cuda:1'), in_proj_covar=tensor([0.0888, 0.0876, 0.0726, 0.0858, 0.0934, 0.0810, 0.0704, 0.0767], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:40:17,294 INFO [train.py:901] (1/4) Epoch 12, batch 7300, loss[loss=0.2292, simple_loss=0.3127, pruned_loss=0.07286, over 8241.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3059, pruned_loss=0.07742, over 1612859.37 frames. ], batch size: 22, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:40:43,441 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.4096, 5.4644, 4.8328, 2.5741, 4.8561, 5.0918, 5.1058, 4.6401], + device='cuda:1'), covar=tensor([0.0576, 0.0407, 0.0836, 0.4182, 0.0715, 0.0835, 0.0952, 0.0775], + device='cuda:1'), in_proj_covar=tensor([0.0467, 0.0380, 0.0389, 0.0480, 0.0381, 0.0383, 0.0378, 0.0332], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:40:45,402 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.567e+02 3.297e+02 4.044e+02 1.170e+03, threshold=6.593e+02, percent-clipped=7.0 +2023-02-06 12:40:51,432 INFO [train.py:901] (1/4) Epoch 12, batch 7350, loss[loss=0.2071, simple_loss=0.2896, pruned_loss=0.06231, over 7655.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.307, pruned_loss=0.07756, over 1613492.27 frames. ], batch size: 19, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:41:09,290 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96289.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:41:15,852 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 12:41:26,504 INFO [train.py:901] (1/4) Epoch 12, batch 7400, loss[loss=0.2365, simple_loss=0.3308, pruned_loss=0.07111, over 8582.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.307, pruned_loss=0.0771, over 1615837.07 frames. ], batch size: 31, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:41:33,425 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:41:36,483 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 12:41:40,268 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8694, 1.6349, 2.1717, 1.8383, 2.0462, 1.7906, 1.5578, 1.1535], + device='cuda:1'), covar=tensor([0.2932, 0.2894, 0.1143, 0.2039, 0.1353, 0.1807, 0.1413, 0.3007], + device='cuda:1'), in_proj_covar=tensor([0.0893, 0.0883, 0.0732, 0.0866, 0.0941, 0.0814, 0.0709, 0.0773], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:41:47,014 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:41:47,084 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:41:52,332 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96350.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:41:55,372 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.734e+02 2.475e+02 3.182e+02 4.307e+02 9.281e+02, threshold=6.365e+02, percent-clipped=3.0 +2023-02-06 12:42:01,559 INFO [train.py:901] (1/4) Epoch 12, batch 7450, loss[loss=0.2389, simple_loss=0.3152, pruned_loss=0.08127, over 7453.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3079, pruned_loss=0.07754, over 1617121.67 frames. ], batch size: 72, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:42:08,570 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2654, 1.2136, 1.4201, 1.1744, 0.6978, 1.2612, 1.1553, 0.9291], + device='cuda:1'), covar=tensor([0.0548, 0.1299, 0.1742, 0.1460, 0.0615, 0.1551, 0.0684, 0.0694], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0152, 0.0191, 0.0157, 0.0103, 0.0162, 0.0115, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 12:42:09,250 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96375.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:42:15,395 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 12:42:31,445 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96408.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:42:35,251 INFO [train.py:901] (1/4) Epoch 12, batch 7500, loss[loss=0.2331, simple_loss=0.3112, pruned_loss=0.07745, over 7812.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3092, pruned_loss=0.07812, over 1617857.48 frames. ], batch size: 20, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:42:54,779 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96442.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:42:57,629 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.3589, 2.8335, 3.7280, 2.2266, 2.0589, 3.5610, 1.0096, 2.1420], + device='cuda:1'), covar=tensor([0.1679, 0.1831, 0.0293, 0.2179, 0.3584, 0.0388, 0.3049, 0.2044], + device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0172, 0.0103, 0.0217, 0.0255, 0.0109, 0.0164, 0.0167], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 12:43:01,568 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9153, 1.4928, 1.6453, 1.3433, 0.8700, 1.3933, 1.5252, 1.5095], + device='cuda:1'), covar=tensor([0.0484, 0.1203, 0.1737, 0.1407, 0.0624, 0.1475, 0.0704, 0.0608], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0151, 0.0190, 0.0157, 0.0103, 0.0161, 0.0115, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 12:43:03,968 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.684e+02 3.354e+02 4.069e+02 8.964e+02, threshold=6.707e+02, percent-clipped=7.0 +2023-02-06 12:43:05,493 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96457.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:43:09,794 INFO [train.py:901] (1/4) Epoch 12, batch 7550, loss[loss=0.2031, simple_loss=0.276, pruned_loss=0.06505, over 7420.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.308, pruned_loss=0.07815, over 1614625.42 frames. ], batch size: 17, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:43:42,989 INFO [train.py:901] (1/4) Epoch 12, batch 7600, loss[loss=0.2493, simple_loss=0.3266, pruned_loss=0.08597, over 8035.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3068, pruned_loss=0.07776, over 1612347.75 frames. ], batch size: 22, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:43:45,372 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.57 vs. limit=5.0 +2023-02-06 12:43:52,179 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-06 12:43:52,500 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96527.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:44:05,570 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96545.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:44:08,120 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3496, 1.3697, 4.5437, 1.8163, 4.0083, 3.7638, 4.0573, 3.9472], + device='cuda:1'), covar=tensor([0.0567, 0.4264, 0.0435, 0.3553, 0.1071, 0.0852, 0.0567, 0.0643], + device='cuda:1'), in_proj_covar=tensor([0.0511, 0.0583, 0.0601, 0.0551, 0.0625, 0.0536, 0.0527, 0.0590], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:44:11,796 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 2.706e+02 3.173e+02 4.121e+02 9.971e+02, threshold=6.345e+02, percent-clipped=8.0 +2023-02-06 12:44:13,329 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96557.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:44:18,536 INFO [train.py:901] (1/4) Epoch 12, batch 7650, loss[loss=0.2188, simple_loss=0.2974, pruned_loss=0.07011, over 8490.00 frames. ], tot_loss[loss=0.2328, simple_loss=0.3086, pruned_loss=0.07853, over 1613442.04 frames. ], batch size: 26, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:44:23,316 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96570.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:44:29,846 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96580.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:44:38,364 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 12:44:47,088 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:44:53,024 INFO [train.py:901] (1/4) Epoch 12, batch 7700, loss[loss=0.2119, simple_loss=0.2985, pruned_loss=0.06261, over 8760.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3069, pruned_loss=0.07743, over 1614323.18 frames. ], batch size: 30, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:45:12,729 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96642.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:45:21,239 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.521e+02 3.004e+02 3.630e+02 7.905e+02, threshold=6.007e+02, percent-clipped=3.0 +2023-02-06 12:45:23,916 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 12:45:27,892 INFO [train.py:901] (1/4) Epoch 12, batch 7750, loss[loss=0.2548, simple_loss=0.3349, pruned_loss=0.08734, over 8565.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3068, pruned_loss=0.07706, over 1618184.30 frames. ], batch size: 39, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:45:34,367 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3675, 1.2574, 2.1164, 1.1681, 1.8282, 2.2714, 2.3527, 1.9731], + device='cuda:1'), covar=tensor([0.0841, 0.1127, 0.0444, 0.1742, 0.0834, 0.0360, 0.0595, 0.0609], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0300, 0.0263, 0.0291, 0.0274, 0.0238, 0.0353, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 12:45:42,858 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96686.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:45:55,113 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9208, 1.5446, 1.6023, 1.3359, 1.0763, 1.3698, 1.6201, 1.4024], + device='cuda:1'), covar=tensor([0.0543, 0.1204, 0.1629, 0.1386, 0.0608, 0.1502, 0.0683, 0.0657], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0151, 0.0190, 0.0156, 0.0103, 0.0161, 0.0114, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 12:46:02,090 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96713.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:46:02,580 INFO [train.py:901] (1/4) Epoch 12, batch 7800, loss[loss=0.2269, simple_loss=0.292, pruned_loss=0.08093, over 6374.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3062, pruned_loss=0.07678, over 1612103.00 frames. ], batch size: 14, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:46:19,269 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96738.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:46:28,440 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96752.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:46:30,315 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 2.519e+02 3.193e+02 4.174e+02 8.059e+02, threshold=6.386e+02, percent-clipped=4.0 +2023-02-06 12:46:36,564 INFO [train.py:901] (1/4) Epoch 12, batch 7850, loss[loss=0.1953, simple_loss=0.27, pruned_loss=0.06034, over 7205.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3069, pruned_loss=0.07731, over 1611157.99 frames. ], batch size: 16, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:46:56,869 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:46:57,907 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-06 12:47:01,901 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96801.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:47:10,007 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:47:10,492 INFO [train.py:901] (1/4) Epoch 12, batch 7900, loss[loss=0.2252, simple_loss=0.3034, pruned_loss=0.07347, over 8356.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3064, pruned_loss=0.07666, over 1608258.30 frames. ], batch size: 26, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:47:27,503 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96838.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:47:38,784 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.595e+02 3.080e+02 3.878e+02 8.124e+02, threshold=6.160e+02, percent-clipped=3.0 +2023-02-06 12:47:44,792 INFO [train.py:901] (1/4) Epoch 12, batch 7950, loss[loss=0.2297, simple_loss=0.2868, pruned_loss=0.08631, over 7253.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.307, pruned_loss=0.07679, over 1607998.78 frames. ], batch size: 16, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:47:47,012 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96867.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:48:07,423 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96898.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:48:17,696 INFO [train.py:901] (1/4) Epoch 12, batch 8000, loss[loss=0.2467, simple_loss=0.3174, pruned_loss=0.08801, over 8424.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3074, pruned_loss=0.07695, over 1611898.77 frames. ], batch size: 49, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:48:23,689 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96923.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:48:45,033 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.377e+02 3.280e+02 4.266e+02 7.100e+02, threshold=6.559e+02, percent-clipped=4.0 +2023-02-06 12:48:51,265 INFO [train.py:901] (1/4) Epoch 12, batch 8050, loss[loss=0.3157, simple_loss=0.374, pruned_loss=0.1287, over 6953.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3082, pruned_loss=0.07811, over 1601951.98 frames. ], batch size: 71, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:49:02,878 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9700, 1.6631, 2.2749, 1.8433, 2.1151, 1.9136, 1.6722, 0.7287], + device='cuda:1'), covar=tensor([0.4466, 0.3799, 0.1307, 0.2450, 0.1832, 0.2325, 0.1656, 0.4032], + device='cuda:1'), in_proj_covar=tensor([0.0896, 0.0884, 0.0734, 0.0859, 0.0939, 0.0816, 0.0708, 0.0773], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:49:10,442 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6471, 2.0184, 2.1953, 1.1788, 2.2828, 1.4570, 0.6506, 1.6910], + device='cuda:1'), covar=tensor([0.0368, 0.0238, 0.0157, 0.0372, 0.0279, 0.0624, 0.0533, 0.0229], + device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0334, 0.0284, 0.0390, 0.0321, 0.0478, 0.0354, 0.0357], + device='cuda:1'), out_proj_covar=tensor([1.1132e-04, 9.1524e-05, 7.8504e-05, 1.0800e-04, 8.9455e-05, 1.4328e-04, + 9.9614e-05, 9.9761e-05], device='cuda:1') +2023-02-06 12:49:24,804 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 12:49:29,797 INFO [train.py:901] (1/4) Epoch 13, batch 0, loss[loss=0.2379, simple_loss=0.3125, pruned_loss=0.08162, over 8341.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3125, pruned_loss=0.08162, over 8341.00 frames. ], batch size: 26, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:49:29,797 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 12:49:40,741 INFO [train.py:935] (1/4) Epoch 13, validation: loss=0.1867, simple_loss=0.2865, pruned_loss=0.04345, over 944034.00 frames. +2023-02-06 12:49:40,743 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 12:49:55,389 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 12:49:55,518 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:50:14,916 INFO [train.py:901] (1/4) Epoch 13, batch 50, loss[loss=0.2667, simple_loss=0.3459, pruned_loss=0.09372, over 8197.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3054, pruned_loss=0.07402, over 365709.23 frames. ], batch size: 23, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:50:20,335 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.833e+02 3.357e+02 4.758e+02 6.927e+02, threshold=6.715e+02, percent-clipped=2.0 +2023-02-06 12:50:21,935 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:50:29,193 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 12:50:41,104 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:50:50,960 INFO [train.py:901] (1/4) Epoch 13, batch 100, loss[loss=0.2084, simple_loss=0.3024, pruned_loss=0.05719, over 8498.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.307, pruned_loss=0.07598, over 646020.17 frames. ], batch size: 26, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:50:52,978 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 12:51:09,078 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97123.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:51:18,982 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97138.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:51:24,703 INFO [train.py:901] (1/4) Epoch 13, batch 150, loss[loss=0.205, simple_loss=0.2876, pruned_loss=0.06124, over 7796.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.306, pruned_loss=0.07629, over 856714.61 frames. ], batch size: 19, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:51:25,593 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97148.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:51:30,112 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.477e+02 2.848e+02 3.342e+02 7.997e+02, threshold=5.696e+02, percent-clipped=2.0 +2023-02-06 12:51:39,278 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-02-06 12:51:42,868 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0684, 1.5326, 3.4335, 1.4503, 2.2630, 3.7885, 3.8459, 3.2620], + device='cuda:1'), covar=tensor([0.1113, 0.1610, 0.0377, 0.2128, 0.1104, 0.0231, 0.0421, 0.0604], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0302, 0.0265, 0.0293, 0.0275, 0.0239, 0.0359, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 12:51:58,452 INFO [train.py:901] (1/4) Epoch 13, batch 200, loss[loss=0.2224, simple_loss=0.3121, pruned_loss=0.06639, over 8242.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3045, pruned_loss=0.07509, over 1026741.77 frames. ], batch size: 24, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:52:09,225 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6979, 3.0661, 2.7601, 4.0833, 1.7982, 2.4634, 2.5107, 3.3851], + device='cuda:1'), covar=tensor([0.0649, 0.0857, 0.0747, 0.0235, 0.1119, 0.1183, 0.1109, 0.0703], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0211, 0.0251, 0.0214, 0.0213, 0.0251, 0.0254, 0.0218], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 12:52:18,266 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5995, 1.4891, 2.8875, 1.0530, 2.0290, 3.0439, 3.3725, 2.2592], + device='cuda:1'), covar=tensor([0.1577, 0.1920, 0.0567, 0.2931, 0.1195, 0.0530, 0.0675, 0.1296], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0301, 0.0263, 0.0292, 0.0274, 0.0238, 0.0356, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 12:52:23,163 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 12:52:33,405 INFO [train.py:901] (1/4) Epoch 13, batch 250, loss[loss=0.1926, simple_loss=0.2756, pruned_loss=0.0548, over 8242.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3053, pruned_loss=0.07525, over 1161598.85 frames. ], batch size: 22, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:52:37,629 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97253.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:52:38,766 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.455e+02 3.117e+02 3.819e+02 7.824e+02, threshold=6.233e+02, percent-clipped=7.0 +2023-02-06 12:52:46,018 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 12:52:53,683 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 12:52:54,017 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:52:54,531 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 12:53:06,366 INFO [train.py:901] (1/4) Epoch 13, batch 300, loss[loss=0.2136, simple_loss=0.2833, pruned_loss=0.07202, over 7525.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3058, pruned_loss=0.07555, over 1264570.46 frames. ], batch size: 18, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:53:06,760 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 12:53:41,550 INFO [train.py:901] (1/4) Epoch 13, batch 350, loss[loss=0.1716, simple_loss=0.2491, pruned_loss=0.04708, over 7643.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3052, pruned_loss=0.07544, over 1340104.42 frames. ], batch size: 19, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:53:46,931 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.701e+02 2.508e+02 3.076e+02 3.709e+02 6.548e+02, threshold=6.153e+02, percent-clipped=1.0 +2023-02-06 12:53:48,446 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1786, 1.4710, 4.3120, 1.5891, 3.7780, 3.6089, 3.8930, 3.7906], + device='cuda:1'), covar=tensor([0.0477, 0.4132, 0.0485, 0.3508, 0.1099, 0.0905, 0.0545, 0.0592], + device='cuda:1'), in_proj_covar=tensor([0.0507, 0.0589, 0.0598, 0.0547, 0.0628, 0.0534, 0.0525, 0.0585], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 12:53:50,467 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97360.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:53:51,766 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97362.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:53:53,796 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97365.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:54:11,875 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:54:15,163 INFO [train.py:901] (1/4) Epoch 13, batch 400, loss[loss=0.2058, simple_loss=0.2885, pruned_loss=0.06159, over 8135.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3044, pruned_loss=0.0748, over 1403635.28 frames. ], batch size: 22, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:54:51,737 INFO [train.py:901] (1/4) Epoch 13, batch 450, loss[loss=0.2347, simple_loss=0.3176, pruned_loss=0.07591, over 8102.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3044, pruned_loss=0.07481, over 1450662.10 frames. ], batch size: 23, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:54:57,109 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.318e+02 2.836e+02 3.756e+02 7.381e+02, threshold=5.672e+02, percent-clipped=3.0 +2023-02-06 12:55:13,276 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:55:26,774 INFO [train.py:901] (1/4) Epoch 13, batch 500, loss[loss=0.2229, simple_loss=0.303, pruned_loss=0.0714, over 8454.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3042, pruned_loss=0.07525, over 1485124.04 frames. ], batch size: 25, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:55:35,518 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97509.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:55:52,996 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97534.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:55:59,606 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 12:56:01,766 INFO [train.py:901] (1/4) Epoch 13, batch 550, loss[loss=0.2364, simple_loss=0.3063, pruned_loss=0.08322, over 8763.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3053, pruned_loss=0.07645, over 1509813.75 frames. ], batch size: 30, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:56:07,720 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.533e+02 3.037e+02 3.770e+02 9.997e+02, threshold=6.074e+02, percent-clipped=4.0 +2023-02-06 12:56:20,096 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:56:36,766 INFO [train.py:901] (1/4) Epoch 13, batch 600, loss[loss=0.2822, simple_loss=0.3538, pruned_loss=0.1052, over 8572.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3059, pruned_loss=0.07597, over 1539026.12 frames. ], batch size: 31, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:56:53,681 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:56:54,431 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7013, 4.7406, 4.2183, 1.9582, 4.1777, 4.3089, 4.4136, 3.9854], + device='cuda:1'), covar=tensor([0.0689, 0.0511, 0.1016, 0.4543, 0.0846, 0.0853, 0.0949, 0.0731], + device='cuda:1'), in_proj_covar=tensor([0.0470, 0.0385, 0.0391, 0.0488, 0.0385, 0.0389, 0.0382, 0.0339], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:56:55,689 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 12:56:55,862 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2567, 1.5568, 1.6484, 1.3752, 1.0807, 1.4538, 1.8450, 1.5105], + device='cuda:1'), covar=tensor([0.0462, 0.1219, 0.1652, 0.1399, 0.0600, 0.1524, 0.0661, 0.0649], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0151, 0.0191, 0.0157, 0.0102, 0.0162, 0.0115, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 12:56:59,743 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3344, 2.1553, 1.7862, 2.0035, 1.6874, 1.3149, 1.5979, 1.7541], + device='cuda:1'), covar=tensor([0.1137, 0.0345, 0.1024, 0.0439, 0.0678, 0.1434, 0.0860, 0.0747], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0238, 0.0324, 0.0304, 0.0306, 0.0327, 0.0345, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 12:57:09,688 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7733, 4.7155, 4.3209, 1.9371, 4.2127, 4.3599, 4.4144, 3.9848], + device='cuda:1'), covar=tensor([0.0568, 0.0534, 0.0951, 0.4707, 0.0765, 0.0653, 0.1082, 0.0716], + device='cuda:1'), in_proj_covar=tensor([0.0467, 0.0385, 0.0391, 0.0487, 0.0384, 0.0388, 0.0380, 0.0338], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 12:57:10,256 INFO [train.py:901] (1/4) Epoch 13, batch 650, loss[loss=0.2236, simple_loss=0.2893, pruned_loss=0.07893, over 7783.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3063, pruned_loss=0.07666, over 1556894.48 frames. ], batch size: 19, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:57:16,274 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 2.537e+02 2.925e+02 3.842e+02 7.324e+02, threshold=5.850e+02, percent-clipped=4.0 +2023-02-06 12:57:37,850 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5792, 2.4813, 1.9398, 2.2151, 2.0932, 1.4170, 1.9332, 2.0805], + device='cuda:1'), covar=tensor([0.1354, 0.0368, 0.1025, 0.0603, 0.0663, 0.1403, 0.0990, 0.0954], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0238, 0.0323, 0.0303, 0.0306, 0.0327, 0.0345, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 12:57:42,540 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97692.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:57:45,736 INFO [train.py:901] (1/4) Epoch 13, batch 700, loss[loss=0.2603, simple_loss=0.3225, pruned_loss=0.0991, over 8133.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3071, pruned_loss=0.07694, over 1575358.20 frames. ], batch size: 22, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:57:50,154 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.26 vs. limit=5.0 +2023-02-06 12:57:51,264 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:57:54,511 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:57:59,099 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-06 12:58:10,714 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:58:12,618 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:58:13,398 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:58:19,712 INFO [train.py:901] (1/4) Epoch 13, batch 750, loss[loss=0.2263, simple_loss=0.3079, pruned_loss=0.07233, over 8583.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3066, pruned_loss=0.07665, over 1587997.74 frames. ], batch size: 34, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:58:25,054 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.478e+02 2.997e+02 3.995e+02 8.399e+02, threshold=5.994e+02, percent-clipped=5.0 +2023-02-06 12:58:27,356 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97758.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:58:39,717 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 12:58:49,019 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 12:58:51,752 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0351, 1.4152, 3.1818, 1.3749, 2.2866, 3.4675, 3.5613, 2.8170], + device='cuda:1'), covar=tensor([0.1084, 0.1751, 0.0478, 0.2220, 0.1078, 0.0345, 0.0593, 0.0825], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0302, 0.0267, 0.0294, 0.0278, 0.0242, 0.0359, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 12:58:54,236 INFO [train.py:901] (1/4) Epoch 13, batch 800, loss[loss=0.2785, simple_loss=0.3475, pruned_loss=0.1047, over 8554.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3063, pruned_loss=0.07626, over 1598738.76 frames. ], batch size: 31, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:59:10,091 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97819.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:59:14,042 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:59:29,483 INFO [train.py:901] (1/4) Epoch 13, batch 850, loss[loss=0.1992, simple_loss=0.285, pruned_loss=0.05669, over 8087.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3067, pruned_loss=0.07661, over 1603028.71 frames. ], batch size: 21, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:59:32,314 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:59:35,504 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.627e+02 3.254e+02 4.246e+02 9.834e+02, threshold=6.507e+02, percent-clipped=8.0 +2023-02-06 13:00:03,794 INFO [train.py:901] (1/4) Epoch 13, batch 900, loss[loss=0.253, simple_loss=0.3261, pruned_loss=0.08993, over 8360.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3057, pruned_loss=0.07625, over 1606974.30 frames. ], batch size: 24, lr: 5.98e-03, grad_scale: 8.0 +2023-02-06 13:00:05,043 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-06 13:00:09,699 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97906.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:00:18,212 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97917.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:00:39,218 INFO [train.py:901] (1/4) Epoch 13, batch 950, loss[loss=0.2685, simple_loss=0.3386, pruned_loss=0.09921, over 8375.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3063, pruned_loss=0.0762, over 1612968.68 frames. ], batch size: 49, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:00:45,307 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.593e+02 3.202e+02 4.020e+02 7.231e+02, threshold=6.403e+02, percent-clipped=2.0 +2023-02-06 13:00:55,387 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9687, 1.6119, 3.3621, 1.3734, 2.3195, 3.7250, 3.7878, 3.1881], + device='cuda:1'), covar=tensor([0.0997, 0.1460, 0.0343, 0.1910, 0.0937, 0.0216, 0.0401, 0.0553], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0300, 0.0266, 0.0292, 0.0276, 0.0241, 0.0358, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:01:08,774 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 13:01:11,143 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:01:13,724 INFO [train.py:901] (1/4) Epoch 13, batch 1000, loss[loss=0.222, simple_loss=0.302, pruned_loss=0.07104, over 8664.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3048, pruned_loss=0.07546, over 1613589.40 frames. ], batch size: 34, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:01:29,783 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:01:39,953 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:01:43,177 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98036.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:01:44,388 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 13:01:50,455 INFO [train.py:901] (1/4) Epoch 13, batch 1050, loss[loss=0.2343, simple_loss=0.2958, pruned_loss=0.0864, over 7796.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3054, pruned_loss=0.07591, over 1615505.98 frames. ], batch size: 19, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:01:56,525 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.707e+02 2.365e+02 2.893e+02 3.782e+02 5.594e+02, threshold=5.785e+02, percent-clipped=0.0 +2023-02-06 13:01:57,223 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 13:02:10,154 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98075.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:13,449 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:24,720 INFO [train.py:901] (1/4) Epoch 13, batch 1100, loss[loss=0.2465, simple_loss=0.3225, pruned_loss=0.08526, over 8572.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3048, pruned_loss=0.07619, over 1610074.54 frames. ], batch size: 31, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:02:26,951 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98100.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:30,280 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98105.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:31,649 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:42,956 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98124.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:49,148 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98132.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:59,481 INFO [train.py:901] (1/4) Epoch 13, batch 1150, loss[loss=0.2387, simple_loss=0.3153, pruned_loss=0.08106, over 8189.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3045, pruned_loss=0.07584, over 1609104.06 frames. ], batch size: 23, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:03:02,367 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98151.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:03:05,425 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 13:03:06,089 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 2.603e+02 3.101e+02 3.825e+02 7.832e+02, threshold=6.203e+02, percent-clipped=6.0 +2023-02-06 13:03:14,947 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1119, 1.2560, 1.2264, 0.8577, 1.3284, 0.9894, 0.3815, 1.1795], + device='cuda:1'), covar=tensor([0.0250, 0.0178, 0.0155, 0.0265, 0.0194, 0.0437, 0.0421, 0.0152], + device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0337, 0.0289, 0.0397, 0.0325, 0.0484, 0.0360, 0.0362], + device='cuda:1'), out_proj_covar=tensor([1.1293e-04, 9.2114e-05, 7.9625e-05, 1.0987e-04, 9.0458e-05, 1.4467e-04, + 1.0148e-04, 1.0106e-04], device='cuda:1') +2023-02-06 13:03:34,163 INFO [train.py:901] (1/4) Epoch 13, batch 1200, loss[loss=0.2497, simple_loss=0.3261, pruned_loss=0.08661, over 8021.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3049, pruned_loss=0.07628, over 1607775.42 frames. ], batch size: 22, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:03:46,029 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-06 13:04:06,233 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2435, 1.0615, 3.3566, 1.0456, 2.9516, 2.8859, 3.0532, 2.9587], + device='cuda:1'), covar=tensor([0.0697, 0.3938, 0.0734, 0.3398, 0.1291, 0.0996, 0.0682, 0.0785], + device='cuda:1'), in_proj_covar=tensor([0.0510, 0.0587, 0.0598, 0.0544, 0.0618, 0.0531, 0.0522, 0.0579], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:04:08,152 INFO [train.py:901] (1/4) Epoch 13, batch 1250, loss[loss=0.2271, simple_loss=0.3227, pruned_loss=0.06575, over 8434.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3048, pruned_loss=0.07586, over 1608157.09 frames. ], batch size: 27, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:04:10,200 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:04:14,074 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.568e+02 3.066e+02 4.053e+02 1.440e+03, threshold=6.132e+02, percent-clipped=8.0 +2023-02-06 13:04:36,960 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:04:43,429 INFO [train.py:901] (1/4) Epoch 13, batch 1300, loss[loss=0.2051, simple_loss=0.2878, pruned_loss=0.06119, over 8029.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3048, pruned_loss=0.07577, over 1608624.04 frames. ], batch size: 22, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:04:44,911 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2304, 1.8111, 2.6039, 2.0623, 2.2358, 2.0553, 1.7356, 1.0606], + device='cuda:1'), covar=tensor([0.3906, 0.3929, 0.1180, 0.2659, 0.1873, 0.2294, 0.1834, 0.4000], + device='cuda:1'), in_proj_covar=tensor([0.0893, 0.0883, 0.0741, 0.0861, 0.0939, 0.0810, 0.0706, 0.0771], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:04:54,306 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6364, 1.4519, 1.6692, 1.2787, 0.9294, 1.4235, 1.5031, 1.4500], + device='cuda:1'), covar=tensor([0.0489, 0.1148, 0.1624, 0.1332, 0.0551, 0.1429, 0.0636, 0.0585], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0153, 0.0193, 0.0158, 0.0102, 0.0164, 0.0116, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 13:04:54,335 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98313.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:05:12,876 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4335, 2.4634, 1.6981, 2.0284, 2.0158, 1.3962, 1.6892, 1.9235], + device='cuda:1'), covar=tensor([0.1226, 0.0312, 0.1010, 0.0569, 0.0573, 0.1306, 0.1009, 0.0830], + device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0231, 0.0313, 0.0295, 0.0296, 0.0317, 0.0334, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:05:16,635 INFO [train.py:901] (1/4) Epoch 13, batch 1350, loss[loss=0.2204, simple_loss=0.3032, pruned_loss=0.06885, over 8328.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3047, pruned_loss=0.07601, over 1605057.28 frames. ], batch size: 25, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:05:23,232 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.492e+02 3.102e+02 3.697e+02 5.327e+02, threshold=6.205e+02, percent-clipped=0.0 +2023-02-06 13:05:29,575 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98365.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:05:52,226 INFO [train.py:901] (1/4) Epoch 13, batch 1400, loss[loss=0.1782, simple_loss=0.2632, pruned_loss=0.04665, over 8255.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3042, pruned_loss=0.07563, over 1611927.09 frames. ], batch size: 22, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:05:59,391 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98407.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:06:14,079 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98428.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:06:16,860 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98432.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:06:26,892 INFO [train.py:901] (1/4) Epoch 13, batch 1450, loss[loss=0.2212, simple_loss=0.3107, pruned_loss=0.06581, over 8599.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3031, pruned_loss=0.07535, over 1609051.74 frames. ], batch size: 39, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:06:32,939 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 2.293e+02 2.812e+02 3.491e+02 8.118e+02, threshold=5.625e+02, percent-clipped=1.0 +2023-02-06 13:06:34,308 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 13:06:41,225 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98468.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:07:02,157 INFO [train.py:901] (1/4) Epoch 13, batch 1500, loss[loss=0.2557, simple_loss=0.3319, pruned_loss=0.08972, over 8463.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3038, pruned_loss=0.0755, over 1610361.55 frames. ], batch size: 25, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:07:37,390 INFO [train.py:901] (1/4) Epoch 13, batch 1550, loss[loss=0.2254, simple_loss=0.2908, pruned_loss=0.07999, over 8078.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3045, pruned_loss=0.07591, over 1611764.50 frames. ], batch size: 21, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:07:38,172 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0560, 1.5278, 1.5846, 1.3291, 0.9612, 1.4228, 1.6470, 1.7042], + device='cuda:1'), covar=tensor([0.0510, 0.1176, 0.1675, 0.1393, 0.0598, 0.1475, 0.0703, 0.0566], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0152, 0.0193, 0.0158, 0.0101, 0.0164, 0.0115, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 13:07:43,379 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.767e+02 2.583e+02 3.208e+02 4.119e+02 6.608e+02, threshold=6.417e+02, percent-clipped=3.0 +2023-02-06 13:07:46,947 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98561.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:08:02,051 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:08:02,732 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5728, 2.7479, 2.3903, 3.7854, 1.8445, 2.0866, 2.2848, 3.0432], + device='cuda:1'), covar=tensor([0.0624, 0.0861, 0.0920, 0.0279, 0.1172, 0.1271, 0.1074, 0.0814], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0216, 0.0258, 0.0217, 0.0218, 0.0255, 0.0260, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 13:08:11,760 INFO [train.py:901] (1/4) Epoch 13, batch 1600, loss[loss=0.2444, simple_loss=0.3218, pruned_loss=0.0835, over 8406.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3053, pruned_loss=0.07568, over 1615019.07 frames. ], batch size: 49, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:08:29,745 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98621.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:08:46,802 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98646.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:08:47,299 INFO [train.py:901] (1/4) Epoch 13, batch 1650, loss[loss=0.1978, simple_loss=0.2653, pruned_loss=0.06517, over 7682.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.305, pruned_loss=0.07575, over 1612747.97 frames. ], batch size: 18, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:08:53,409 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.462e+02 2.942e+02 3.707e+02 8.113e+02, threshold=5.885e+02, percent-clipped=6.0 +2023-02-06 13:09:20,787 INFO [train.py:901] (1/4) Epoch 13, batch 1700, loss[loss=0.2105, simple_loss=0.2906, pruned_loss=0.06519, over 8233.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3047, pruned_loss=0.07544, over 1611568.28 frames. ], batch size: 22, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:09:43,783 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5140, 1.9151, 3.3350, 1.2615, 2.6463, 1.9881, 1.6898, 2.4109], + device='cuda:1'), covar=tensor([0.1907, 0.2346, 0.0826, 0.4274, 0.1529, 0.3082, 0.1946, 0.2149], + device='cuda:1'), in_proj_covar=tensor([0.0491, 0.0532, 0.0537, 0.0589, 0.0620, 0.0558, 0.0480, 0.0611], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 13:09:57,380 INFO [train.py:901] (1/4) Epoch 13, batch 1750, loss[loss=0.2504, simple_loss=0.3322, pruned_loss=0.08426, over 8349.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3054, pruned_loss=0.07581, over 1615893.49 frames. ], batch size: 49, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:10:03,434 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.663e+02 3.311e+02 3.905e+02 7.561e+02, threshold=6.622e+02, percent-clipped=6.0 +2023-02-06 13:10:15,122 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98772.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:10:31,882 INFO [train.py:901] (1/4) Epoch 13, batch 1800, loss[loss=0.2377, simple_loss=0.317, pruned_loss=0.07918, over 8363.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.307, pruned_loss=0.07619, over 1622126.54 frames. ], batch size: 24, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:11:01,386 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98839.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:11:06,994 INFO [train.py:901] (1/4) Epoch 13, batch 1850, loss[loss=0.2914, simple_loss=0.3452, pruned_loss=0.1188, over 8424.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3061, pruned_loss=0.07623, over 1620890.62 frames. ], batch size: 49, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:11:13,499 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.419e+02 2.914e+02 4.067e+02 1.078e+03, threshold=5.828e+02, percent-clipped=2.0 +2023-02-06 13:11:18,918 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98864.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:11:29,406 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9071, 3.7546, 2.3484, 2.3793, 2.5815, 1.8056, 2.5621, 2.8087], + device='cuda:1'), covar=tensor([0.1463, 0.0254, 0.0917, 0.0809, 0.0718, 0.1353, 0.1023, 0.0966], + device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0235, 0.0316, 0.0299, 0.0300, 0.0323, 0.0339, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:11:34,565 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98887.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:11:41,190 INFO [train.py:901] (1/4) Epoch 13, batch 1900, loss[loss=0.2035, simple_loss=0.2839, pruned_loss=0.06158, over 7913.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3063, pruned_loss=0.07623, over 1622617.78 frames. ], batch size: 20, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:11:46,605 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98905.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:11:48,677 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98908.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:11:51,825 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98913.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:12:11,195 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 13:12:15,140 INFO [train.py:901] (1/4) Epoch 13, batch 1950, loss[loss=0.2603, simple_loss=0.3252, pruned_loss=0.09768, over 8511.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3058, pruned_loss=0.07608, over 1620715.74 frames. ], batch size: 28, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:12:21,305 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.445e+02 3.079e+02 3.874e+02 6.986e+02, threshold=6.158e+02, percent-clipped=4.0 +2023-02-06 13:12:23,955 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 13:12:35,732 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8576, 6.0156, 5.2939, 2.2904, 5.2068, 5.6608, 5.4730, 5.3107], + device='cuda:1'), covar=tensor([0.0523, 0.0432, 0.1004, 0.4617, 0.0790, 0.0600, 0.1156, 0.0598], + device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0390, 0.0396, 0.0492, 0.0388, 0.0392, 0.0384, 0.0341], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 13:12:44,576 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 13:12:45,368 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98989.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:12:50,409 INFO [train.py:901] (1/4) Epoch 13, batch 2000, loss[loss=0.2884, simple_loss=0.3733, pruned_loss=0.1018, over 8509.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3053, pruned_loss=0.07588, over 1619563.95 frames. ], batch size: 28, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:13:06,451 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99020.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:13:24,249 INFO [train.py:901] (1/4) Epoch 13, batch 2050, loss[loss=0.1625, simple_loss=0.247, pruned_loss=0.03905, over 7777.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3048, pruned_loss=0.07553, over 1621519.96 frames. ], batch size: 19, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:13:30,114 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.527e+02 3.267e+02 4.166e+02 9.227e+02, threshold=6.535e+02, percent-clipped=8.0 +2023-02-06 13:13:39,427 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99069.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:13:43,693 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 13:13:51,522 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7806, 1.2459, 3.9569, 1.4861, 3.4350, 3.2797, 3.5475, 3.4404], + device='cuda:1'), covar=tensor([0.0641, 0.4479, 0.0532, 0.3661, 0.1265, 0.0990, 0.0593, 0.0737], + device='cuda:1'), in_proj_covar=tensor([0.0511, 0.0581, 0.0594, 0.0550, 0.0624, 0.0534, 0.0524, 0.0587], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:13:52,406 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-02-06 13:13:58,777 INFO [train.py:901] (1/4) Epoch 13, batch 2100, loss[loss=0.2389, simple_loss=0.3091, pruned_loss=0.08433, over 7810.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3036, pruned_loss=0.07473, over 1619847.49 frames. ], batch size: 20, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:14:27,699 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7977, 1.4048, 3.9457, 1.3590, 3.4719, 3.2435, 3.5106, 3.4316], + device='cuda:1'), covar=tensor([0.0657, 0.4496, 0.0672, 0.4119, 0.1365, 0.1081, 0.0711, 0.0826], + device='cuda:1'), in_proj_covar=tensor([0.0512, 0.0583, 0.0596, 0.0550, 0.0624, 0.0534, 0.0523, 0.0587], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:14:31,280 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99143.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:14:33,893 INFO [train.py:901] (1/4) Epoch 13, batch 2150, loss[loss=0.2107, simple_loss=0.2802, pruned_loss=0.07054, over 8085.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3038, pruned_loss=0.075, over 1614257.41 frames. ], batch size: 21, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:14:39,911 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.393e+02 2.767e+02 3.323e+02 5.467e+02, threshold=5.533e+02, percent-clipped=0.0 +2023-02-06 13:14:48,054 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99168.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:15:07,845 INFO [train.py:901] (1/4) Epoch 13, batch 2200, loss[loss=0.1806, simple_loss=0.2647, pruned_loss=0.04826, over 7214.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3046, pruned_loss=0.07565, over 1615355.26 frames. ], batch size: 16, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:15:27,754 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2325, 1.9308, 3.2589, 1.5694, 2.4772, 3.5777, 3.5621, 3.1157], + device='cuda:1'), covar=tensor([0.0900, 0.1283, 0.0430, 0.1946, 0.1027, 0.0230, 0.0538, 0.0552], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0303, 0.0270, 0.0295, 0.0280, 0.0241, 0.0361, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:15:43,035 INFO [train.py:901] (1/4) Epoch 13, batch 2250, loss[loss=0.2418, simple_loss=0.3047, pruned_loss=0.08947, over 7799.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3048, pruned_loss=0.07592, over 1612527.17 frames. ], batch size: 19, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:15:43,151 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:15:46,453 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99252.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:15:48,894 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 2.623e+02 3.377e+02 4.135e+02 6.545e+02, threshold=6.753e+02, percent-clipped=6.0 +2023-02-06 13:15:49,684 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99257.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:02,580 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99276.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:09,204 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:16,503 INFO [train.py:901] (1/4) Epoch 13, batch 2300, loss[loss=0.2556, simple_loss=0.3314, pruned_loss=0.08994, over 8353.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3041, pruned_loss=0.07535, over 1610168.13 frames. ], batch size: 24, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:16:19,363 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99301.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:43,128 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99333.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:49,178 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:52,411 INFO [train.py:901] (1/4) Epoch 13, batch 2350, loss[loss=0.242, simple_loss=0.3133, pruned_loss=0.08536, over 6819.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3045, pruned_loss=0.07585, over 1607979.31 frames. ], batch size: 72, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:16:58,394 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.470e+02 3.074e+02 3.865e+02 1.080e+03, threshold=6.149e+02, percent-clipped=3.0 +2023-02-06 13:17:06,771 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99367.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:17:10,175 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99372.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:17:12,140 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7105, 1.2048, 4.8687, 1.7214, 4.4060, 4.0991, 4.4309, 4.3047], + device='cuda:1'), covar=tensor([0.0452, 0.4597, 0.0375, 0.3723, 0.0807, 0.0791, 0.0448, 0.0525], + device='cuda:1'), in_proj_covar=tensor([0.0521, 0.0596, 0.0613, 0.0564, 0.0637, 0.0546, 0.0538, 0.0601], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:17:22,902 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0478, 2.2707, 2.3184, 1.6219, 2.4942, 1.8237, 1.6199, 1.9720], + device='cuda:1'), covar=tensor([0.0510, 0.0243, 0.0168, 0.0431, 0.0251, 0.0422, 0.0485, 0.0328], + device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0336, 0.0292, 0.0402, 0.0325, 0.0487, 0.0361, 0.0367], + device='cuda:1'), out_proj_covar=tensor([1.1254e-04, 9.1758e-05, 8.0184e-05, 1.1101e-04, 9.0365e-05, 1.4522e-04, + 1.0168e-04, 1.0207e-04], device='cuda:1') +2023-02-06 13:17:26,585 INFO [train.py:901] (1/4) Epoch 13, batch 2400, loss[loss=0.2158, simple_loss=0.2966, pruned_loss=0.06752, over 8653.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3047, pruned_loss=0.07602, over 1606819.57 frames. ], batch size: 39, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:17:37,521 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99413.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:17:54,055 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2877, 1.9325, 2.7580, 2.2242, 2.5611, 2.1927, 1.8130, 1.2952], + device='cuda:1'), covar=tensor([0.4237, 0.4344, 0.1464, 0.2830, 0.2047, 0.2490, 0.1746, 0.4409], + device='cuda:1'), in_proj_covar=tensor([0.0896, 0.0887, 0.0740, 0.0862, 0.0942, 0.0813, 0.0705, 0.0775], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:17:55,981 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:18:01,835 INFO [train.py:901] (1/4) Epoch 13, batch 2450, loss[loss=0.2415, simple_loss=0.3171, pruned_loss=0.08299, over 8515.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3037, pruned_loss=0.07577, over 1601171.97 frames. ], batch size: 29, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:18:02,707 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99448.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:18:08,609 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.541e+02 3.157e+02 3.793e+02 6.756e+02, threshold=6.314e+02, percent-clipped=3.0 +2023-02-06 13:18:29,563 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99486.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:18:36,515 INFO [train.py:901] (1/4) Epoch 13, batch 2500, loss[loss=0.2706, simple_loss=0.339, pruned_loss=0.1011, over 7450.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3032, pruned_loss=0.07529, over 1601783.31 frames. ], batch size: 72, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:18:45,256 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9704, 2.7284, 3.6738, 2.2787, 1.9078, 3.5297, 0.7671, 2.2027], + device='cuda:1'), covar=tensor([0.2114, 0.1348, 0.0239, 0.2351, 0.3698, 0.0484, 0.3327, 0.1783], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0176, 0.0104, 0.0220, 0.0257, 0.0111, 0.0165, 0.0167], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 13:18:57,692 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99528.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:19:10,748 INFO [train.py:901] (1/4) Epoch 13, batch 2550, loss[loss=0.2458, simple_loss=0.3102, pruned_loss=0.0907, over 7707.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3029, pruned_loss=0.0752, over 1602494.57 frames. ], batch size: 18, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:19:17,199 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.420e+02 2.977e+02 3.875e+02 7.325e+02, threshold=5.954e+02, percent-clipped=4.0 +2023-02-06 13:19:19,406 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6568, 2.3317, 3.4490, 2.6668, 2.9815, 2.3741, 2.0144, 1.8526], + device='cuda:1'), covar=tensor([0.3872, 0.4153, 0.1357, 0.2838, 0.2156, 0.2372, 0.1631, 0.4603], + device='cuda:1'), in_proj_covar=tensor([0.0893, 0.0887, 0.0738, 0.0862, 0.0943, 0.0812, 0.0703, 0.0774], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:19:20,014 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7018, 1.8258, 2.5468, 1.6769, 1.1482, 2.4366, 0.3956, 1.3671], + device='cuda:1'), covar=tensor([0.2653, 0.1592, 0.0412, 0.2378, 0.4636, 0.0496, 0.3524, 0.2097], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0176, 0.0103, 0.0219, 0.0257, 0.0110, 0.0165, 0.0167], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 13:19:21,127 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6653, 2.6483, 2.0053, 2.2204, 2.2717, 1.6541, 1.9396, 2.2283], + device='cuda:1'), covar=tensor([0.1229, 0.0340, 0.0826, 0.0526, 0.0551, 0.1147, 0.0903, 0.0724], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0240, 0.0320, 0.0300, 0.0301, 0.0326, 0.0344, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:19:37,733 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99586.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:19:41,078 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99591.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:19:45,635 INFO [train.py:901] (1/4) Epoch 13, batch 2600, loss[loss=0.2591, simple_loss=0.3272, pruned_loss=0.09551, over 8695.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3041, pruned_loss=0.07546, over 1611186.03 frames. ], batch size: 49, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:19:52,602 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:03,510 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:06,924 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99628.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:08,162 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:19,551 INFO [train.py:901] (1/4) Epoch 13, batch 2650, loss[loss=0.251, simple_loss=0.3376, pruned_loss=0.08222, over 8556.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3043, pruned_loss=0.07509, over 1614300.66 frames. ], batch size: 31, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:20:20,458 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99648.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:23,636 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99653.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:25,434 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.403e+02 3.099e+02 4.031e+02 8.160e+02, threshold=6.198e+02, percent-clipped=1.0 +2023-02-06 13:20:47,382 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99686.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:54,454 INFO [train.py:901] (1/4) Epoch 13, batch 2700, loss[loss=0.2154, simple_loss=0.2933, pruned_loss=0.06868, over 7659.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3045, pruned_loss=0.07487, over 1618536.19 frames. ], batch size: 19, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:20:55,210 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:59,153 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8291, 1.4967, 1.8011, 1.4774, 1.1183, 1.6579, 2.2025, 2.1767], + device='cuda:1'), covar=tensor([0.0420, 0.1358, 0.1793, 0.1462, 0.0591, 0.1561, 0.0625, 0.0603], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0192, 0.0158, 0.0102, 0.0163, 0.0115, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 13:20:59,201 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99704.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:00,475 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99706.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:12,257 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 13:21:16,796 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99729.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:27,694 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:28,869 INFO [train.py:901] (1/4) Epoch 13, batch 2750, loss[loss=0.2268, simple_loss=0.3099, pruned_loss=0.07186, over 8254.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3041, pruned_loss=0.07468, over 1616024.89 frames. ], batch size: 24, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:21:34,776 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.508e+02 3.194e+02 3.866e+02 8.318e+02, threshold=6.387e+02, percent-clipped=3.0 +2023-02-06 13:21:52,456 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99782.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:53,978 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99784.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:55,002 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 13:21:57,553 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 13:22:03,191 INFO [train.py:901] (1/4) Epoch 13, batch 2800, loss[loss=0.2386, simple_loss=0.3263, pruned_loss=0.07544, over 8491.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3055, pruned_loss=0.07557, over 1617059.67 frames. ], batch size: 26, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:22:06,170 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99801.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:22:11,449 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99809.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:22:26,105 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99830.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:22:37,606 INFO [train.py:901] (1/4) Epoch 13, batch 2850, loss[loss=0.2106, simple_loss=0.2831, pruned_loss=0.06908, over 7662.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3046, pruned_loss=0.0753, over 1615877.49 frames. ], batch size: 19, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:22:43,877 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.486e+02 2.909e+02 3.673e+02 9.445e+02, threshold=5.818e+02, percent-clipped=3.0 +2023-02-06 13:23:11,478 INFO [train.py:901] (1/4) Epoch 13, batch 2900, loss[loss=0.2204, simple_loss=0.3029, pruned_loss=0.06896, over 8034.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.305, pruned_loss=0.07579, over 1612949.00 frames. ], batch size: 22, lr: 5.92e-03, grad_scale: 16.0 +2023-02-06 13:23:11,680 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99897.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:23:34,894 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99930.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:23:44,996 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99945.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:23:46,141 INFO [train.py:901] (1/4) Epoch 13, batch 2950, loss[loss=0.221, simple_loss=0.2833, pruned_loss=0.07937, over 7971.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3055, pruned_loss=0.0763, over 1612364.57 frames. ], batch size: 21, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:23:48,968 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:23:52,246 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 13:23:52,911 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.762e+02 3.280e+02 4.150e+02 8.176e+02, threshold=6.560e+02, percent-clipped=12.0 +2023-02-06 13:23:57,303 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99962.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:24:14,501 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99987.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:24:21,176 INFO [train.py:901] (1/4) Epoch 13, batch 3000, loss[loss=0.2194, simple_loss=0.2899, pruned_loss=0.07452, over 7801.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3047, pruned_loss=0.07555, over 1608323.02 frames. ], batch size: 19, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:24:21,176 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 13:24:33,566 INFO [train.py:935] (1/4) Epoch 13, validation: loss=0.1841, simple_loss=0.2841, pruned_loss=0.04204, over 944034.00 frames. +2023-02-06 13:24:33,567 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 13:24:37,722 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100001.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:24:42,513 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3255, 1.1799, 1.4386, 1.1919, 0.7342, 1.2745, 1.2878, 1.1196], + device='cuda:1'), covar=tensor([0.0514, 0.1362, 0.1843, 0.1444, 0.0600, 0.1568, 0.0681, 0.0657], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0192, 0.0158, 0.0102, 0.0163, 0.0115, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 13:24:53,937 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100025.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:24:55,300 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100026.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:05,693 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:07,828 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:09,025 INFO [train.py:901] (1/4) Epoch 13, batch 3050, loss[loss=0.2242, simple_loss=0.31, pruned_loss=0.06926, over 8337.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3055, pruned_loss=0.07588, over 1613194.37 frames. ], batch size: 26, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:25:15,849 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.522e+02 3.008e+02 4.207e+02 1.157e+03, threshold=6.017e+02, percent-clipped=6.0 +2023-02-06 13:25:16,792 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:22,900 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:24,259 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1501, 1.2132, 4.2759, 1.6369, 3.7506, 3.5111, 3.8548, 3.7766], + device='cuda:1'), covar=tensor([0.0526, 0.4481, 0.0520, 0.3641, 0.1145, 0.1065, 0.0578, 0.0620], + device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0591, 0.0609, 0.0558, 0.0634, 0.0546, 0.0538, 0.0591], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:25:34,609 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:41,469 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9925, 2.7930, 3.4789, 2.0803, 1.8598, 3.5279, 0.8726, 2.2527], + device='cuda:1'), covar=tensor([0.2226, 0.1445, 0.0330, 0.2569, 0.3965, 0.0384, 0.3299, 0.1836], + device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0177, 0.0106, 0.0222, 0.0260, 0.0112, 0.0166, 0.0170], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 13:25:44,733 INFO [train.py:901] (1/4) Epoch 13, batch 3100, loss[loss=0.2624, simple_loss=0.3409, pruned_loss=0.09196, over 8359.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3062, pruned_loss=0.07614, over 1616132.83 frames. ], batch size: 24, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:26:19,849 INFO [train.py:901] (1/4) Epoch 13, batch 3150, loss[loss=0.1893, simple_loss=0.27, pruned_loss=0.05433, over 8136.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3055, pruned_loss=0.07551, over 1618001.84 frames. ], batch size: 22, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:26:24,151 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:26:25,932 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.491e+02 3.036e+02 4.077e+02 6.258e+02, threshold=6.072e+02, percent-clipped=1.0 +2023-02-06 13:26:26,823 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100157.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:26:41,745 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100178.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:26:55,151 INFO [train.py:901] (1/4) Epoch 13, batch 3200, loss[loss=0.2024, simple_loss=0.2877, pruned_loss=0.05854, over 7810.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3046, pruned_loss=0.07509, over 1617233.31 frames. ], batch size: 20, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:26:58,247 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100201.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:27:05,241 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0660, 1.4868, 1.5837, 1.3782, 1.0188, 1.4238, 1.7094, 1.8365], + device='cuda:1'), covar=tensor([0.0527, 0.1281, 0.1692, 0.1421, 0.0624, 0.1527, 0.0711, 0.0554], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0191, 0.0158, 0.0102, 0.0163, 0.0114, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 13:27:15,349 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100226.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:27:29,408 INFO [train.py:901] (1/4) Epoch 13, batch 3250, loss[loss=0.2488, simple_loss=0.3222, pruned_loss=0.08772, over 8588.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3049, pruned_loss=0.07537, over 1620426.27 frames. ], batch size: 34, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:27:34,918 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8355, 3.7877, 3.4199, 1.7765, 3.4084, 3.3881, 3.4026, 3.1134], + device='cuda:1'), covar=tensor([0.0889, 0.0684, 0.1109, 0.4137, 0.0926, 0.0968, 0.1400, 0.1002], + device='cuda:1'), in_proj_covar=tensor([0.0475, 0.0387, 0.0391, 0.0490, 0.0388, 0.0392, 0.0385, 0.0342], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 13:27:35,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.430e+02 2.991e+02 3.670e+02 7.489e+02, threshold=5.982e+02, percent-clipped=4.0 +2023-02-06 13:27:35,627 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100256.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:28:04,522 INFO [train.py:901] (1/4) Epoch 13, batch 3300, loss[loss=0.2601, simple_loss=0.3337, pruned_loss=0.09323, over 8517.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3047, pruned_loss=0.07543, over 1613530.46 frames. ], batch size: 26, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:28:07,523 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100301.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:28:22,245 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:28:24,889 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100326.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:28:39,548 INFO [train.py:901] (1/4) Epoch 13, batch 3350, loss[loss=0.1937, simple_loss=0.261, pruned_loss=0.06317, over 7687.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3039, pruned_loss=0.07471, over 1614180.78 frames. ], batch size: 18, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:28:39,750 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:28:45,562 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.599e+02 3.166e+02 3.997e+02 7.990e+02, threshold=6.333e+02, percent-clipped=6.0 +2023-02-06 13:28:54,364 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100369.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:29:13,503 INFO [train.py:901] (1/4) Epoch 13, batch 3400, loss[loss=0.1871, simple_loss=0.2598, pruned_loss=0.05724, over 7700.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3034, pruned_loss=0.07453, over 1613578.92 frames. ], batch size: 18, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:29:25,268 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100413.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:29:29,914 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100420.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:29:42,785 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:29:48,792 INFO [train.py:901] (1/4) Epoch 13, batch 3450, loss[loss=0.2711, simple_loss=0.3302, pruned_loss=0.106, over 8532.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3016, pruned_loss=0.0731, over 1611156.41 frames. ], batch size: 39, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:29:54,821 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.523e+02 3.011e+02 4.006e+02 7.808e+02, threshold=6.023e+02, percent-clipped=2.0 +2023-02-06 13:30:14,314 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100484.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:30:23,422 INFO [train.py:901] (1/4) Epoch 13, batch 3500, loss[loss=0.223, simple_loss=0.2955, pruned_loss=0.07522, over 7929.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3027, pruned_loss=0.07398, over 1611297.78 frames. ], batch size: 20, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:30:26,341 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9739, 2.4599, 3.6357, 2.7233, 3.1012, 2.6523, 2.2589, 2.0314], + device='cuda:1'), covar=tensor([0.3395, 0.3947, 0.1231, 0.2707, 0.2167, 0.2050, 0.1534, 0.4277], + device='cuda:1'), in_proj_covar=tensor([0.0882, 0.0881, 0.0727, 0.0849, 0.0932, 0.0808, 0.0699, 0.0768], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:30:53,031 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 13:30:58,419 INFO [train.py:901] (1/4) Epoch 13, batch 3550, loss[loss=0.2131, simple_loss=0.2914, pruned_loss=0.06741, over 8462.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3031, pruned_loss=0.07453, over 1601823.82 frames. ], batch size: 25, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:31:04,446 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.564e+02 3.091e+02 3.906e+02 9.185e+02, threshold=6.182e+02, percent-clipped=3.0 +2023-02-06 13:31:33,167 INFO [train.py:901] (1/4) Epoch 13, batch 3600, loss[loss=0.3194, simple_loss=0.3685, pruned_loss=0.1351, over 7071.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3025, pruned_loss=0.07439, over 1601058.99 frames. ], batch size: 71, lr: 5.89e-03, grad_scale: 16.0 +2023-02-06 13:31:35,333 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100600.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:31:40,865 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2402, 1.5470, 1.6757, 1.3951, 1.0997, 1.4745, 1.9573, 1.9142], + device='cuda:1'), covar=tensor([0.0471, 0.1188, 0.1645, 0.1316, 0.0603, 0.1463, 0.0602, 0.0569], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0151, 0.0189, 0.0156, 0.0101, 0.0162, 0.0114, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 13:32:08,079 INFO [train.py:901] (1/4) Epoch 13, batch 3650, loss[loss=0.2569, simple_loss=0.333, pruned_loss=0.09044, over 8265.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3025, pruned_loss=0.07465, over 1599469.78 frames. ], batch size: 24, lr: 5.89e-03, grad_scale: 16.0 +2023-02-06 13:32:12,473 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 +2023-02-06 13:32:14,110 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.362e+02 3.080e+02 3.827e+02 7.938e+02, threshold=6.161e+02, percent-clipped=3.0 +2023-02-06 13:32:38,642 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5753, 1.3319, 4.7282, 1.7449, 4.1825, 3.9566, 4.2862, 4.1359], + device='cuda:1'), covar=tensor([0.0476, 0.4562, 0.0397, 0.3429, 0.0904, 0.0774, 0.0490, 0.0542], + device='cuda:1'), in_proj_covar=tensor([0.0518, 0.0599, 0.0616, 0.0560, 0.0637, 0.0548, 0.0537, 0.0597], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:32:43,156 INFO [train.py:901] (1/4) Epoch 13, batch 3700, loss[loss=0.216, simple_loss=0.2953, pruned_loss=0.06836, over 8449.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3019, pruned_loss=0.07444, over 1601517.47 frames. ], batch size: 27, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:32:55,265 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:32:57,153 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 13:33:09,354 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-06 13:33:12,128 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-02-06 13:33:13,302 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100740.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:33:17,859 INFO [train.py:901] (1/4) Epoch 13, batch 3750, loss[loss=0.2563, simple_loss=0.3297, pruned_loss=0.09143, over 8116.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3025, pruned_loss=0.0745, over 1603076.16 frames. ], batch size: 23, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:33:18,692 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100748.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:33:24,692 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.543e+02 3.029e+02 3.909e+02 6.778e+02, threshold=6.059e+02, percent-clipped=2.0 +2023-02-06 13:33:28,257 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5287, 1.9717, 2.2477, 1.1123, 2.3287, 1.3222, 0.6795, 1.6998], + device='cuda:1'), covar=tensor([0.0582, 0.0273, 0.0207, 0.0525, 0.0268, 0.0743, 0.0735, 0.0319], + device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0341, 0.0290, 0.0398, 0.0327, 0.0489, 0.0363, 0.0367], + device='cuda:1'), out_proj_covar=tensor([1.1185e-04, 9.3302e-05, 7.9554e-05, 1.0965e-04, 9.0508e-05, 1.4573e-04, + 1.0217e-04, 1.0191e-04], device='cuda:1') +2023-02-06 13:33:30,145 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100764.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:33:31,007 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100765.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:33:52,963 INFO [train.py:901] (1/4) Epoch 13, batch 3800, loss[loss=0.197, simple_loss=0.2634, pruned_loss=0.06532, over 7766.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.303, pruned_loss=0.0746, over 1606565.21 frames. ], batch size: 19, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:34:22,774 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 13:34:27,591 INFO [train.py:901] (1/4) Epoch 13, batch 3850, loss[loss=0.2445, simple_loss=0.3256, pruned_loss=0.08174, over 8546.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3031, pruned_loss=0.07465, over 1604893.09 frames. ], batch size: 31, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:34:34,506 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.985e+02 2.830e+02 3.312e+02 3.730e+02 7.453e+02, threshold=6.624e+02, percent-clipped=3.0 +2023-02-06 13:34:50,102 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100879.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:35:01,746 INFO [train.py:901] (1/4) Epoch 13, batch 3900, loss[loss=0.2024, simple_loss=0.2889, pruned_loss=0.05792, over 8279.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3045, pruned_loss=0.07545, over 1610436.54 frames. ], batch size: 23, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:35:01,756 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 13:35:12,828 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 13:35:18,420 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.46 vs. limit=5.0 +2023-02-06 13:35:33,879 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0037, 1.8444, 4.2279, 1.7330, 2.4702, 4.6102, 4.7737, 3.8435], + device='cuda:1'), covar=tensor([0.1345, 0.1627, 0.0296, 0.2259, 0.1123, 0.0287, 0.0423, 0.0819], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0301, 0.0265, 0.0293, 0.0277, 0.0239, 0.0359, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:35:37,127 INFO [train.py:901] (1/4) Epoch 13, batch 3950, loss[loss=0.2227, simple_loss=0.3075, pruned_loss=0.06891, over 8241.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3047, pruned_loss=0.0753, over 1612580.56 frames. ], batch size: 22, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:35:44,007 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.489e+02 3.011e+02 3.855e+02 9.802e+02, threshold=6.021e+02, percent-clipped=2.0 +2023-02-06 13:35:54,143 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100971.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:36:11,866 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100996.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:36:12,368 INFO [train.py:901] (1/4) Epoch 13, batch 4000, loss[loss=0.2516, simple_loss=0.3282, pruned_loss=0.08747, over 8614.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3061, pruned_loss=0.07585, over 1616804.36 frames. ], batch size: 49, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:36:47,681 INFO [train.py:901] (1/4) Epoch 13, batch 4050, loss[loss=0.2241, simple_loss=0.3104, pruned_loss=0.06896, over 8285.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3062, pruned_loss=0.07571, over 1618891.09 frames. ], batch size: 23, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:36:54,252 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.767e+02 2.645e+02 3.184e+02 3.816e+02 9.518e+02, threshold=6.368e+02, percent-clipped=3.0 +2023-02-06 13:37:17,940 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-02-06 13:37:18,351 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101092.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:37:18,940 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5885, 4.6206, 4.1408, 1.9556, 4.1253, 4.1394, 4.1955, 3.9284], + device='cuda:1'), covar=tensor([0.0884, 0.0613, 0.1204, 0.4843, 0.0834, 0.0861, 0.1400, 0.0810], + device='cuda:1'), in_proj_covar=tensor([0.0480, 0.0388, 0.0399, 0.0497, 0.0392, 0.0393, 0.0387, 0.0341], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 13:37:21,159 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6837, 1.9810, 2.1112, 1.2703, 2.2508, 1.4824, 0.7440, 1.8374], + device='cuda:1'), covar=tensor([0.0408, 0.0198, 0.0159, 0.0384, 0.0229, 0.0562, 0.0542, 0.0193], + device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0338, 0.0290, 0.0400, 0.0328, 0.0487, 0.0366, 0.0367], + device='cuda:1'), out_proj_covar=tensor([1.1169e-04, 9.2409e-05, 7.9217e-05, 1.1059e-04, 9.0656e-05, 1.4486e-04, + 1.0278e-04, 1.0186e-04], device='cuda:1') +2023-02-06 13:37:21,621 INFO [train.py:901] (1/4) Epoch 13, batch 4100, loss[loss=0.2614, simple_loss=0.3347, pruned_loss=0.09401, over 8553.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3052, pruned_loss=0.07514, over 1618150.50 frames. ], batch size: 39, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:37:41,405 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:37:47,943 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:37:50,602 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5662, 1.5299, 4.6953, 1.6682, 4.1096, 3.8707, 4.2396, 4.1275], + device='cuda:1'), covar=tensor([0.0470, 0.4480, 0.0434, 0.3774, 0.1085, 0.0932, 0.0540, 0.0610], + device='cuda:1'), in_proj_covar=tensor([0.0515, 0.0591, 0.0619, 0.0556, 0.0634, 0.0547, 0.0536, 0.0600], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:37:55,653 INFO [train.py:901] (1/4) Epoch 13, batch 4150, loss[loss=0.2419, simple_loss=0.3433, pruned_loss=0.07023, over 8102.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3067, pruned_loss=0.07632, over 1615040.43 frames. ], batch size: 23, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:38:02,999 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.783e+02 3.401e+02 4.642e+02 1.010e+03, threshold=6.803e+02, percent-clipped=7.0 +2023-02-06 13:38:05,271 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101160.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:38:30,233 INFO [train.py:901] (1/4) Epoch 13, batch 4200, loss[loss=0.2025, simple_loss=0.2856, pruned_loss=0.05974, over 8103.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3072, pruned_loss=0.07676, over 1613507.26 frames. ], batch size: 23, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:38:37,994 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101207.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:38:54,601 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 13:39:05,853 INFO [train.py:901] (1/4) Epoch 13, batch 4250, loss[loss=0.2179, simple_loss=0.2945, pruned_loss=0.07067, over 8233.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3069, pruned_loss=0.07665, over 1616776.37 frames. ], batch size: 22, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:39:12,480 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.514e+02 3.154e+02 3.992e+02 7.648e+02, threshold=6.307e+02, percent-clipped=3.0 +2023-02-06 13:39:16,514 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 13:39:40,007 INFO [train.py:901] (1/4) Epoch 13, batch 4300, loss[loss=0.1951, simple_loss=0.2648, pruned_loss=0.06266, over 7429.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3068, pruned_loss=0.07657, over 1617547.18 frames. ], batch size: 17, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:40:14,576 INFO [train.py:901] (1/4) Epoch 13, batch 4350, loss[loss=0.1988, simple_loss=0.2637, pruned_loss=0.06697, over 7234.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3059, pruned_loss=0.07599, over 1616840.93 frames. ], batch size: 16, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:40:21,344 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.682e+02 3.184e+02 4.441e+02 9.358e+02, threshold=6.368e+02, percent-clipped=11.0 +2023-02-06 13:40:30,790 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7122, 1.3794, 3.8754, 1.4683, 3.3945, 3.2270, 3.5065, 3.4218], + device='cuda:1'), covar=tensor([0.0588, 0.4064, 0.0652, 0.3562, 0.1120, 0.0888, 0.0577, 0.0648], + device='cuda:1'), in_proj_covar=tensor([0.0510, 0.0584, 0.0615, 0.0551, 0.0623, 0.0537, 0.0531, 0.0590], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:40:49,428 INFO [train.py:901] (1/4) Epoch 13, batch 4400, loss[loss=0.2435, simple_loss=0.317, pruned_loss=0.08498, over 8449.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3068, pruned_loss=0.07598, over 1620716.89 frames. ], batch size: 29, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:40:49,436 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 13:41:20,991 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9981, 2.2396, 1.9292, 2.7911, 1.0694, 1.5558, 1.8177, 2.3005], + device='cuda:1'), covar=tensor([0.0774, 0.0855, 0.0940, 0.0413, 0.1257, 0.1405, 0.1022, 0.0800], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0212, 0.0256, 0.0214, 0.0216, 0.0254, 0.0260, 0.0219], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 13:41:22,953 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9950, 1.7311, 3.4029, 1.4127, 2.2329, 3.7390, 3.8077, 3.1647], + device='cuda:1'), covar=tensor([0.1079, 0.1522, 0.0307, 0.2173, 0.1039, 0.0211, 0.0437, 0.0585], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0302, 0.0266, 0.0296, 0.0278, 0.0240, 0.0360, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:41:23,440 INFO [train.py:901] (1/4) Epoch 13, batch 4450, loss[loss=0.1956, simple_loss=0.2745, pruned_loss=0.05836, over 7800.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3069, pruned_loss=0.07587, over 1619634.38 frames. ], batch size: 20, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:41:28,826 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 13:41:30,664 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 2.700e+02 3.319e+02 4.103e+02 1.285e+03, threshold=6.638e+02, percent-clipped=3.0 +2023-02-06 13:41:34,860 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101463.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:41:38,376 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-06 13:41:38,578 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101469.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:41:52,078 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101488.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:41:57,888 INFO [train.py:901] (1/4) Epoch 13, batch 4500, loss[loss=0.2623, simple_loss=0.3277, pruned_loss=0.09846, over 7968.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.307, pruned_loss=0.07582, over 1618552.51 frames. ], batch size: 21, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:42:04,780 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-06 13:42:22,190 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 13:42:30,588 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1010, 1.7739, 2.4624, 1.9131, 2.3322, 1.9547, 1.6812, 1.1381], + device='cuda:1'), covar=tensor([0.3723, 0.3733, 0.1318, 0.2657, 0.1656, 0.2371, 0.1705, 0.3769], + device='cuda:1'), in_proj_covar=tensor([0.0893, 0.0893, 0.0746, 0.0868, 0.0944, 0.0818, 0.0709, 0.0778], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:42:33,108 INFO [train.py:901] (1/4) Epoch 13, batch 4550, loss[loss=0.2338, simple_loss=0.3127, pruned_loss=0.07746, over 8143.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3062, pruned_loss=0.07546, over 1616549.19 frames. ], batch size: 22, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:42:39,101 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-02-06 13:42:39,884 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 2.403e+02 2.986e+02 3.546e+02 6.918e+02, threshold=5.973e+02, percent-clipped=1.0 +2023-02-06 13:42:58,931 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101584.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:43:08,196 INFO [train.py:901] (1/4) Epoch 13, batch 4600, loss[loss=0.2487, simple_loss=0.3289, pruned_loss=0.08421, over 8467.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.305, pruned_loss=0.07443, over 1619643.98 frames. ], batch size: 49, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:43:42,608 INFO [train.py:901] (1/4) Epoch 13, batch 4650, loss[loss=0.2009, simple_loss=0.282, pruned_loss=0.05989, over 7542.00 frames. ], tot_loss[loss=0.227, simple_loss=0.305, pruned_loss=0.07451, over 1620848.55 frames. ], batch size: 18, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:43:49,456 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 2.500e+02 2.989e+02 3.844e+02 7.619e+02, threshold=5.978e+02, percent-clipped=4.0 +2023-02-06 13:44:00,410 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2830, 1.5657, 4.4836, 1.9236, 3.9358, 3.7098, 4.0370, 3.9016], + device='cuda:1'), covar=tensor([0.0557, 0.4384, 0.0467, 0.3311, 0.0992, 0.0781, 0.0524, 0.0607], + device='cuda:1'), in_proj_covar=tensor([0.0510, 0.0581, 0.0613, 0.0547, 0.0622, 0.0534, 0.0526, 0.0589], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:44:17,686 INFO [train.py:901] (1/4) Epoch 13, batch 4700, loss[loss=0.2153, simple_loss=0.2865, pruned_loss=0.0721, over 7659.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3044, pruned_loss=0.07443, over 1620067.93 frames. ], batch size: 19, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:44:21,754 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101702.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:44:52,087 INFO [train.py:901] (1/4) Epoch 13, batch 4750, loss[loss=0.2067, simple_loss=0.2809, pruned_loss=0.0662, over 7791.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3027, pruned_loss=0.07372, over 1614314.35 frames. ], batch size: 19, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:44:55,560 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0982, 1.6370, 3.3015, 1.5041, 2.2859, 3.6123, 3.7212, 3.0602], + device='cuda:1'), covar=tensor([0.0993, 0.1484, 0.0315, 0.1975, 0.0992, 0.0238, 0.0499, 0.0593], + device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0299, 0.0264, 0.0294, 0.0276, 0.0237, 0.0358, 0.0294], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:44:59,502 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.558e+02 3.081e+02 3.778e+02 8.564e+02, threshold=6.162e+02, percent-clipped=2.0 +2023-02-06 13:45:21,975 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 13:45:24,390 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 13:45:27,102 INFO [train.py:901] (1/4) Epoch 13, batch 4800, loss[loss=0.2252, simple_loss=0.3017, pruned_loss=0.07435, over 8370.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3026, pruned_loss=0.07381, over 1613240.01 frames. ], batch size: 48, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:45:57,593 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101840.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:46:01,954 INFO [train.py:901] (1/4) Epoch 13, batch 4850, loss[loss=0.2447, simple_loss=0.3291, pruned_loss=0.08019, over 8331.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3031, pruned_loss=0.07427, over 1610170.91 frames. ], batch size: 26, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:46:08,646 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.589e+02 3.137e+02 3.918e+02 7.572e+02, threshold=6.274e+02, percent-clipped=4.0 +2023-02-06 13:46:14,101 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 13:46:14,946 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101865.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:46:36,888 INFO [train.py:901] (1/4) Epoch 13, batch 4900, loss[loss=0.2033, simple_loss=0.2997, pruned_loss=0.05349, over 8453.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3036, pruned_loss=0.07418, over 1614912.18 frames. ], batch size: 29, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:46:45,576 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-06 13:47:06,381 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101938.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:47:12,292 INFO [train.py:901] (1/4) Epoch 13, batch 4950, loss[loss=0.2376, simple_loss=0.3254, pruned_loss=0.07493, over 8327.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3031, pruned_loss=0.07393, over 1609966.69 frames. ], batch size: 25, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:47:19,104 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.579e+02 2.432e+02 3.023e+02 3.670e+02 7.494e+02, threshold=6.046e+02, percent-clipped=3.0 +2023-02-06 13:47:30,006 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6959, 1.4555, 1.6040, 1.3173, 0.8994, 1.3466, 1.4601, 1.4330], + device='cuda:1'), covar=tensor([0.0491, 0.1202, 0.1619, 0.1338, 0.0567, 0.1507, 0.0687, 0.0597], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0191, 0.0157, 0.0102, 0.0163, 0.0114, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 13:47:46,750 INFO [train.py:901] (1/4) Epoch 13, batch 5000, loss[loss=0.2282, simple_loss=0.3085, pruned_loss=0.07397, over 8337.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3021, pruned_loss=0.07363, over 1607733.29 frames. ], batch size: 26, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:47:53,355 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:48:16,611 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0180, 1.6559, 1.7420, 1.6447, 1.1955, 1.8377, 2.1206, 2.0976], + device='cuda:1'), covar=tensor([0.0391, 0.1214, 0.1634, 0.1321, 0.0583, 0.1429, 0.0613, 0.0534], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0191, 0.0158, 0.0102, 0.0163, 0.0114, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 13:48:22,481 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102046.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:48:23,084 INFO [train.py:901] (1/4) Epoch 13, batch 5050, loss[loss=0.1864, simple_loss=0.2631, pruned_loss=0.05485, over 8038.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3016, pruned_loss=0.07382, over 1606529.74 frames. ], batch size: 20, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:48:29,968 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.626e+02 3.300e+02 4.185e+02 9.088e+02, threshold=6.599e+02, percent-clipped=3.0 +2023-02-06 13:48:54,021 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 13:48:56,649 INFO [train.py:901] (1/4) Epoch 13, batch 5100, loss[loss=0.2246, simple_loss=0.3131, pruned_loss=0.068, over 7964.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3035, pruned_loss=0.07449, over 1604956.06 frames. ], batch size: 21, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:49:31,561 INFO [train.py:901] (1/4) Epoch 13, batch 5150, loss[loss=0.1911, simple_loss=0.2693, pruned_loss=0.05643, over 8240.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3048, pruned_loss=0.0755, over 1608391.44 frames. ], batch size: 22, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:49:38,306 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.413e+02 2.853e+02 3.425e+02 7.647e+02, threshold=5.706e+02, percent-clipped=3.0 +2023-02-06 13:49:41,794 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:50:06,648 INFO [train.py:901] (1/4) Epoch 13, batch 5200, loss[loss=0.2242, simple_loss=0.307, pruned_loss=0.07067, over 8474.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3042, pruned_loss=0.07553, over 1606617.81 frames. ], batch size: 25, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:50:41,853 INFO [train.py:901] (1/4) Epoch 13, batch 5250, loss[loss=0.2391, simple_loss=0.3277, pruned_loss=0.07527, over 8496.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3046, pruned_loss=0.07538, over 1608237.64 frames. ], batch size: 29, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:50:48,575 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.565e+02 3.047e+02 3.925e+02 1.157e+03, threshold=6.094e+02, percent-clipped=6.0 +2023-02-06 13:50:51,018 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.79 vs. limit=5.0 +2023-02-06 13:50:53,901 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 13:51:06,839 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102282.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:51:16,478 INFO [train.py:901] (1/4) Epoch 13, batch 5300, loss[loss=0.2034, simple_loss=0.2863, pruned_loss=0.06025, over 8602.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3033, pruned_loss=0.07469, over 1606391.99 frames. ], batch size: 31, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:51:51,008 INFO [train.py:901] (1/4) Epoch 13, batch 5350, loss[loss=0.2477, simple_loss=0.3251, pruned_loss=0.08515, over 8506.00 frames. ], tot_loss[loss=0.227, simple_loss=0.304, pruned_loss=0.075, over 1611509.84 frames. ], batch size: 26, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:51:52,483 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:51:52,614 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5955, 2.7796, 1.7947, 2.1871, 2.2138, 1.4933, 2.0296, 2.1930], + device='cuda:1'), covar=tensor([0.1474, 0.0395, 0.1234, 0.0707, 0.0700, 0.1493, 0.1075, 0.1090], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0236, 0.0318, 0.0297, 0.0298, 0.0324, 0.0345, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:51:57,799 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.535e+02 3.049e+02 3.805e+02 7.372e+02, threshold=6.098e+02, percent-clipped=2.0 +2023-02-06 13:52:08,205 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 +2023-02-06 13:52:26,065 INFO [train.py:901] (1/4) Epoch 13, batch 5400, loss[loss=0.2384, simple_loss=0.3119, pruned_loss=0.08246, over 8743.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3048, pruned_loss=0.07536, over 1616189.70 frames. ], batch size: 39, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:52:26,257 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102397.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:52:40,305 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:52:56,882 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102442.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:53:00,140 INFO [train.py:901] (1/4) Epoch 13, batch 5450, loss[loss=0.2169, simple_loss=0.3041, pruned_loss=0.06489, over 8356.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3056, pruned_loss=0.0758, over 1614639.34 frames. ], batch size: 24, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:53:07,658 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.724e+02 3.222e+02 3.900e+02 7.023e+02, threshold=6.444e+02, percent-clipped=3.0 +2023-02-06 13:53:12,594 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102464.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:53:34,976 INFO [train.py:901] (1/4) Epoch 13, batch 5500, loss[loss=0.2024, simple_loss=0.2786, pruned_loss=0.06313, over 7534.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3049, pruned_loss=0.07549, over 1612773.96 frames. ], batch size: 18, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:53:41,586 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 13:54:03,099 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0110, 1.7628, 3.3235, 1.3184, 2.2711, 3.7226, 3.8154, 3.1283], + device='cuda:1'), covar=tensor([0.1117, 0.1473, 0.0350, 0.2328, 0.1014, 0.0214, 0.0397, 0.0597], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0300, 0.0265, 0.0296, 0.0277, 0.0237, 0.0360, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:54:09,472 INFO [train.py:901] (1/4) Epoch 13, batch 5550, loss[loss=0.1678, simple_loss=0.2497, pruned_loss=0.04293, over 7228.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3045, pruned_loss=0.0761, over 1608001.59 frames. ], batch size: 16, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:54:15,943 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 2.279e+02 3.010e+02 3.933e+02 6.976e+02, threshold=6.019e+02, percent-clipped=1.0 +2023-02-06 13:54:43,196 INFO [train.py:901] (1/4) Epoch 13, batch 5600, loss[loss=0.202, simple_loss=0.2717, pruned_loss=0.06609, over 7810.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3042, pruned_loss=0.07542, over 1611624.29 frames. ], batch size: 19, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:54:44,043 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3003, 1.6131, 4.3682, 2.0847, 2.3727, 5.0583, 5.0747, 4.3232], + device='cuda:1'), covar=tensor([0.1103, 0.1651, 0.0273, 0.1939, 0.1112, 0.0195, 0.0346, 0.0583], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0302, 0.0266, 0.0296, 0.0278, 0.0239, 0.0360, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 13:55:04,663 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2566, 2.5714, 2.9102, 1.4749, 2.9047, 1.8874, 1.5811, 2.1077], + device='cuda:1'), covar=tensor([0.0575, 0.0269, 0.0213, 0.0555, 0.0414, 0.0574, 0.0650, 0.0361], + device='cuda:1'), in_proj_covar=tensor([0.0406, 0.0341, 0.0293, 0.0401, 0.0332, 0.0490, 0.0368, 0.0373], + device='cuda:1'), out_proj_covar=tensor([1.1347e-04, 9.3025e-05, 7.9922e-05, 1.1031e-04, 9.1562e-05, 1.4528e-04, + 1.0327e-04, 1.0324e-04], device='cuda:1') +2023-02-06 13:55:18,285 INFO [train.py:901] (1/4) Epoch 13, batch 5650, loss[loss=0.2189, simple_loss=0.2924, pruned_loss=0.07267, over 8240.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3051, pruned_loss=0.07575, over 1611632.84 frames. ], batch size: 22, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:55:22,473 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102653.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:55:24,858 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.730e+02 3.267e+02 4.266e+02 8.129e+02, threshold=6.534e+02, percent-clipped=5.0 +2023-02-06 13:55:39,201 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102678.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:55:43,646 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 13:55:52,506 INFO [train.py:901] (1/4) Epoch 13, batch 5700, loss[loss=0.2263, simple_loss=0.2986, pruned_loss=0.07698, over 8563.00 frames. ], tot_loss[loss=0.227, simple_loss=0.304, pruned_loss=0.07501, over 1614597.29 frames. ], batch size: 39, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:56:08,740 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102720.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:56:25,991 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:56:27,755 INFO [train.py:901] (1/4) Epoch 13, batch 5750, loss[loss=0.2056, simple_loss=0.2856, pruned_loss=0.06282, over 8246.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.304, pruned_loss=0.07441, over 1614126.42 frames. ], batch size: 22, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:56:34,417 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.889e+02 2.514e+02 3.075e+02 4.012e+02 7.214e+02, threshold=6.150e+02, percent-clipped=2.0 +2023-02-06 13:56:45,826 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.77 vs. limit=5.0 +2023-02-06 13:56:47,305 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 13:56:49,609 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5441, 2.1768, 3.2814, 2.5676, 2.8602, 2.3339, 2.0730, 1.7960], + device='cuda:1'), covar=tensor([0.3959, 0.4336, 0.1515, 0.3168, 0.2553, 0.2521, 0.1696, 0.4821], + device='cuda:1'), in_proj_covar=tensor([0.0894, 0.0889, 0.0744, 0.0870, 0.0943, 0.0815, 0.0705, 0.0778], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:56:52,946 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8150, 2.1034, 2.3131, 1.2134, 2.4206, 1.6840, 0.6765, 2.0057], + device='cuda:1'), covar=tensor([0.0477, 0.0243, 0.0181, 0.0509, 0.0243, 0.0633, 0.0680, 0.0229], + device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0342, 0.0295, 0.0403, 0.0334, 0.0492, 0.0369, 0.0374], + device='cuda:1'), out_proj_covar=tensor([1.1393e-04, 9.3301e-05, 8.0550e-05, 1.1094e-04, 9.2128e-05, 1.4583e-04, + 1.0362e-04, 1.0363e-04], device='cuda:1') +2023-02-06 13:57:01,390 INFO [train.py:901] (1/4) Epoch 13, batch 5800, loss[loss=0.2379, simple_loss=0.2899, pruned_loss=0.09301, over 7252.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3035, pruned_loss=0.07399, over 1615695.62 frames. ], batch size: 16, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:57:36,617 INFO [train.py:901] (1/4) Epoch 13, batch 5850, loss[loss=0.2137, simple_loss=0.2853, pruned_loss=0.07099, over 7800.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3026, pruned_loss=0.07367, over 1615994.23 frames. ], batch size: 20, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:57:43,168 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.218e+02 2.874e+02 3.517e+02 7.476e+02, threshold=5.748e+02, percent-clipped=3.0 +2023-02-06 13:57:52,589 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102869.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:58:04,221 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6090, 1.7163, 2.2394, 1.4188, 1.1717, 2.2070, 0.3776, 1.3992], + device='cuda:1'), covar=tensor([0.2813, 0.1626, 0.0422, 0.1848, 0.3834, 0.0432, 0.3001, 0.1763], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0174, 0.0106, 0.0219, 0.0255, 0.0110, 0.0164, 0.0171], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 13:58:11,326 INFO [train.py:901] (1/4) Epoch 13, batch 5900, loss[loss=0.2817, simple_loss=0.3372, pruned_loss=0.1131, over 8122.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.303, pruned_loss=0.07399, over 1617921.81 frames. ], batch size: 22, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:58:16,843 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102905.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:58:17,966 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-02-06 13:58:46,019 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 13:58:46,265 INFO [train.py:901] (1/4) Epoch 13, batch 5950, loss[loss=0.2378, simple_loss=0.3144, pruned_loss=0.08062, over 8021.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3029, pruned_loss=0.07396, over 1616709.15 frames. ], batch size: 22, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:58:52,891 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.606e+02 2.537e+02 3.124e+02 4.010e+02 1.248e+03, threshold=6.247e+02, percent-clipped=9.0 +2023-02-06 13:59:01,045 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6139, 1.6691, 2.2862, 1.5704, 0.9962, 2.1909, 0.4121, 1.3873], + device='cuda:1'), covar=tensor([0.2508, 0.1563, 0.0447, 0.1791, 0.3808, 0.0491, 0.2935, 0.1725], + device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0175, 0.0107, 0.0221, 0.0256, 0.0111, 0.0166, 0.0172], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 13:59:21,490 INFO [train.py:901] (1/4) Epoch 13, batch 6000, loss[loss=0.1955, simple_loss=0.2855, pruned_loss=0.0528, over 8233.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3041, pruned_loss=0.07478, over 1621243.01 frames. ], batch size: 24, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:59:21,490 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 13:59:30,390 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8459, 1.0161, 2.9810, 1.0743, 2.5896, 2.4539, 2.6255, 2.5370], + device='cuda:1'), covar=tensor([0.0472, 0.3475, 0.0383, 0.3206, 0.0910, 0.0749, 0.0471, 0.0618], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0581, 0.0604, 0.0552, 0.0624, 0.0529, 0.0524, 0.0582], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 13:59:36,606 INFO [train.py:935] (1/4) Epoch 13, validation: loss=0.1836, simple_loss=0.2836, pruned_loss=0.04176, over 944034.00 frames. +2023-02-06 13:59:36,607 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 14:00:11,030 INFO [train.py:901] (1/4) Epoch 13, batch 6050, loss[loss=0.192, simple_loss=0.2732, pruned_loss=0.05543, over 7797.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3024, pruned_loss=0.07382, over 1617379.92 frames. ], batch size: 19, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:00:18,306 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.701e+02 2.480e+02 3.014e+02 3.999e+02 8.436e+02, threshold=6.027e+02, percent-clipped=4.0 +2023-02-06 14:00:25,996 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6739, 1.8056, 1.6763, 2.3433, 1.1820, 1.4549, 1.6940, 1.9414], + device='cuda:1'), covar=tensor([0.0721, 0.0802, 0.0899, 0.0407, 0.1011, 0.1274, 0.0735, 0.0668], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0209, 0.0252, 0.0211, 0.0216, 0.0251, 0.0256, 0.0215], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:00:45,637 INFO [train.py:901] (1/4) Epoch 13, batch 6100, loss[loss=0.2628, simple_loss=0.3391, pruned_loss=0.09328, over 8367.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.304, pruned_loss=0.07437, over 1617006.31 frames. ], batch size: 49, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:00:58,586 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103116.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:01:14,163 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 14:01:19,578 INFO [train.py:901] (1/4) Epoch 13, batch 6150, loss[loss=0.1956, simple_loss=0.2728, pruned_loss=0.05921, over 7929.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3036, pruned_loss=0.07443, over 1618252.01 frames. ], batch size: 20, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:01:21,559 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103150.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:01:26,205 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.434e+02 3.117e+02 4.172e+02 7.466e+02, threshold=6.235e+02, percent-clipped=2.0 +2023-02-06 14:01:46,186 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6754, 1.5788, 2.2026, 1.2340, 1.8259, 2.4129, 2.3704, 2.1247], + device='cuda:1'), covar=tensor([0.0798, 0.1082, 0.0641, 0.1697, 0.1230, 0.0308, 0.0780, 0.0547], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0303, 0.0265, 0.0297, 0.0278, 0.0239, 0.0361, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:01:51,050 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4909, 1.8152, 4.4264, 1.9393, 2.3672, 5.0597, 5.0814, 4.4041], + device='cuda:1'), covar=tensor([0.1038, 0.1511, 0.0280, 0.1918, 0.1177, 0.0168, 0.0404, 0.0528], + device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0301, 0.0264, 0.0296, 0.0277, 0.0239, 0.0359, 0.0294], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:01:55,081 INFO [train.py:901] (1/4) Epoch 13, batch 6200, loss[loss=0.2595, simple_loss=0.3376, pruned_loss=0.09066, over 8508.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3043, pruned_loss=0.07516, over 1617944.44 frames. ], batch size: 26, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:02:06,145 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:02:31,003 INFO [train.py:901] (1/4) Epoch 13, batch 6250, loss[loss=0.2751, simple_loss=0.3394, pruned_loss=0.1054, over 6622.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3032, pruned_loss=0.07501, over 1611738.03 frames. ], batch size: 71, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:02:32,368 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103249.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:02:37,850 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.482e+02 2.950e+02 3.630e+02 6.819e+02, threshold=5.900e+02, percent-clipped=4.0 +2023-02-06 14:03:06,048 INFO [train.py:901] (1/4) Epoch 13, batch 6300, loss[loss=0.2327, simple_loss=0.3137, pruned_loss=0.07586, over 8094.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3036, pruned_loss=0.07461, over 1615179.96 frames. ], batch size: 21, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:03:27,932 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:03:40,543 INFO [train.py:901] (1/4) Epoch 13, batch 6350, loss[loss=0.1863, simple_loss=0.2658, pruned_loss=0.05336, over 7777.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.305, pruned_loss=0.0757, over 1616704.81 frames. ], batch size: 19, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:03:48,232 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.547e+02 3.093e+02 3.716e+02 8.603e+02, threshold=6.185e+02, percent-clipped=3.0 +2023-02-06 14:03:52,991 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:04:02,165 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3838, 4.3382, 3.9687, 1.6962, 3.9419, 3.9485, 4.0660, 3.5600], + device='cuda:1'), covar=tensor([0.0793, 0.0595, 0.1055, 0.5221, 0.0844, 0.0877, 0.1118, 0.1003], + device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0386, 0.0395, 0.0491, 0.0390, 0.0388, 0.0380, 0.0340], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 14:04:06,343 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103384.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:04:11,086 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9697, 2.2509, 1.8889, 2.7645, 1.2699, 1.5918, 1.8914, 2.2349], + device='cuda:1'), covar=tensor([0.0737, 0.0738, 0.0880, 0.0337, 0.1207, 0.1409, 0.0917, 0.0682], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0208, 0.0252, 0.0212, 0.0216, 0.0252, 0.0256, 0.0215], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:04:14,877 INFO [train.py:901] (1/4) Epoch 13, batch 6400, loss[loss=0.2598, simple_loss=0.3157, pruned_loss=0.102, over 7919.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3051, pruned_loss=0.07579, over 1613534.99 frames. ], batch size: 20, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:04:36,360 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7991, 1.9951, 2.3125, 1.5292, 2.3556, 1.6402, 0.8105, 1.9605], + device='cuda:1'), covar=tensor([0.0443, 0.0243, 0.0176, 0.0367, 0.0292, 0.0595, 0.0588, 0.0216], + device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0341, 0.0293, 0.0402, 0.0333, 0.0491, 0.0365, 0.0370], + device='cuda:1'), out_proj_covar=tensor([1.1384e-04, 9.3002e-05, 7.9762e-05, 1.1053e-04, 9.1891e-05, 1.4560e-04, + 1.0235e-04, 1.0221e-04], device='cuda:1') +2023-02-06 14:04:49,452 INFO [train.py:901] (1/4) Epoch 13, batch 6450, loss[loss=0.3192, simple_loss=0.3779, pruned_loss=0.1303, over 7096.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.304, pruned_loss=0.07567, over 1609893.69 frames. ], batch size: 74, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:04:56,176 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.890e+02 2.528e+02 3.186e+02 3.863e+02 6.544e+02, threshold=6.372e+02, percent-clipped=1.0 +2023-02-06 14:04:58,239 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103460.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:05:22,249 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103494.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:05:24,240 INFO [train.py:901] (1/4) Epoch 13, batch 6500, loss[loss=0.2008, simple_loss=0.2896, pruned_loss=0.05599, over 8331.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3037, pruned_loss=0.07462, over 1614106.17 frames. ], batch size: 25, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:05:30,977 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6463, 1.9122, 2.1609, 1.3052, 2.2481, 1.4486, 0.6415, 1.8771], + device='cuda:1'), covar=tensor([0.0497, 0.0298, 0.0207, 0.0447, 0.0284, 0.0738, 0.0666, 0.0220], + device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0343, 0.0295, 0.0404, 0.0333, 0.0494, 0.0367, 0.0372], + device='cuda:1'), out_proj_covar=tensor([1.1472e-04, 9.3462e-05, 8.0153e-05, 1.1115e-04, 9.1904e-05, 1.4647e-04, + 1.0306e-04, 1.0281e-04], device='cuda:1') +2023-02-06 14:05:32,164 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6819, 1.3071, 1.5687, 1.2443, 0.9085, 1.3607, 1.4652, 1.3977], + device='cuda:1'), covar=tensor([0.0542, 0.1308, 0.1700, 0.1481, 0.0595, 0.1513, 0.0698, 0.0645], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0150, 0.0189, 0.0155, 0.0100, 0.0161, 0.0113, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 14:05:58,666 INFO [train.py:901] (1/4) Epoch 13, batch 6550, loss[loss=0.176, simple_loss=0.2592, pruned_loss=0.04644, over 7975.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3041, pruned_loss=0.07448, over 1616088.10 frames. ], batch size: 21, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:06:05,488 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.442e+02 3.089e+02 4.027e+02 9.292e+02, threshold=6.177e+02, percent-clipped=8.0 +2023-02-06 14:06:18,304 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103575.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:06:24,184 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 14:06:24,389 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103584.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:06:34,121 INFO [train.py:901] (1/4) Epoch 13, batch 6600, loss[loss=0.1976, simple_loss=0.2703, pruned_loss=0.06246, over 7530.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3032, pruned_loss=0.07426, over 1613306.88 frames. ], batch size: 18, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:06:42,481 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103609.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:06:42,507 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103609.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:06:43,685 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 14:06:49,823 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:07:07,869 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:07:09,058 INFO [train.py:901] (1/4) Epoch 13, batch 6650, loss[loss=0.2411, simple_loss=0.3093, pruned_loss=0.08645, over 7812.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.304, pruned_loss=0.07457, over 1611676.88 frames. ], batch size: 20, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:07:16,597 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.459e+02 2.800e+02 3.637e+02 6.016e+02, threshold=5.600e+02, percent-clipped=0.0 +2023-02-06 14:07:38,061 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4329, 2.6167, 1.8429, 2.1709, 2.1553, 1.4925, 1.9926, 2.0275], + device='cuda:1'), covar=tensor([0.1396, 0.0331, 0.0973, 0.0662, 0.0681, 0.1426, 0.0897, 0.0991], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0236, 0.0321, 0.0298, 0.0298, 0.0326, 0.0343, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:07:43,986 INFO [train.py:901] (1/4) Epoch 13, batch 6700, loss[loss=0.2771, simple_loss=0.3324, pruned_loss=0.111, over 6817.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3035, pruned_loss=0.07424, over 1610721.83 frames. ], batch size: 15, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:07:54,502 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9846, 2.3485, 1.8198, 3.0278, 1.3705, 1.7007, 2.0000, 2.3629], + device='cuda:1'), covar=tensor([0.0800, 0.0851, 0.1047, 0.0338, 0.1225, 0.1428, 0.0921, 0.0801], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0205, 0.0250, 0.0209, 0.0214, 0.0249, 0.0253, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:08:06,482 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103728.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:08:19,449 INFO [train.py:901] (1/4) Epoch 13, batch 6750, loss[loss=0.2439, simple_loss=0.3249, pruned_loss=0.08142, over 8475.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3035, pruned_loss=0.07436, over 1613693.00 frames. ], batch size: 28, lr: 5.80e-03, grad_scale: 16.0 +2023-02-06 14:08:26,130 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.551e+02 3.234e+02 3.983e+02 1.044e+03, threshold=6.469e+02, percent-clipped=6.0 +2023-02-06 14:08:26,993 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103758.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:08:44,329 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103783.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:08:53,553 INFO [train.py:901] (1/4) Epoch 13, batch 6800, loss[loss=0.2687, simple_loss=0.3525, pruned_loss=0.09242, over 8487.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3054, pruned_loss=0.07581, over 1613003.67 frames. ], batch size: 28, lr: 5.80e-03, grad_scale: 16.0 +2023-02-06 14:08:56,499 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9889, 2.1443, 1.7376, 2.5871, 1.2581, 1.6266, 1.7631, 2.1174], + device='cuda:1'), covar=tensor([0.0696, 0.0778, 0.0886, 0.0475, 0.1165, 0.1227, 0.0922, 0.0829], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0205, 0.0249, 0.0208, 0.0212, 0.0248, 0.0252, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:08:58,345 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 14:09:17,173 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103831.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:09:25,225 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103843.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 14:09:27,548 INFO [train.py:901] (1/4) Epoch 13, batch 6850, loss[loss=0.2452, simple_loss=0.3357, pruned_loss=0.07735, over 8605.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.305, pruned_loss=0.07521, over 1612265.08 frames. ], batch size: 31, lr: 5.80e-03, grad_scale: 16.0 +2023-02-06 14:09:33,737 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:09:34,148 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.955e+02 2.670e+02 3.153e+02 3.957e+02 9.275e+02, threshold=6.306e+02, percent-clipped=2.0 +2023-02-06 14:09:40,247 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103865.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:09:44,812 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 14:09:51,041 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0260, 4.0290, 2.7405, 2.9161, 2.9276, 2.3365, 3.0129, 3.2046], + device='cuda:1'), covar=tensor([0.1687, 0.0352, 0.0834, 0.0684, 0.0656, 0.1236, 0.0898, 0.1066], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0234, 0.0318, 0.0295, 0.0298, 0.0324, 0.0340, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:09:57,527 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103890.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:10:00,233 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103894.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:10:02,103 INFO [train.py:901] (1/4) Epoch 13, batch 6900, loss[loss=0.1866, simple_loss=0.2741, pruned_loss=0.04961, over 7809.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3046, pruned_loss=0.07514, over 1610967.94 frames. ], batch size: 20, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:10:35,828 INFO [train.py:901] (1/4) Epoch 13, batch 6950, loss[loss=0.184, simple_loss=0.2558, pruned_loss=0.05606, over 7265.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3037, pruned_loss=0.07428, over 1612360.16 frames. ], batch size: 16, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:10:43,024 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.539e+02 3.074e+02 3.917e+02 9.810e+02, threshold=6.147e+02, percent-clipped=9.0 +2023-02-06 14:10:53,178 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 14:11:09,724 INFO [train.py:901] (1/4) Epoch 13, batch 7000, loss[loss=0.2328, simple_loss=0.3136, pruned_loss=0.07606, over 8291.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.303, pruned_loss=0.07413, over 1611955.82 frames. ], batch size: 23, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:11:18,192 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104008.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:11:23,553 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104015.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:11:39,557 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8401, 1.3894, 5.9069, 2.0986, 5.3379, 5.0311, 5.4906, 5.3996], + device='cuda:1'), covar=tensor([0.0474, 0.4770, 0.0379, 0.3469, 0.0944, 0.0711, 0.0417, 0.0520], + device='cuda:1'), in_proj_covar=tensor([0.0522, 0.0589, 0.0613, 0.0560, 0.0635, 0.0538, 0.0530, 0.0595], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 14:11:44,820 INFO [train.py:901] (1/4) Epoch 13, batch 7050, loss[loss=0.2341, simple_loss=0.3156, pruned_loss=0.07626, over 8458.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3036, pruned_loss=0.07452, over 1611346.67 frames. ], batch size: 27, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:11:51,571 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1868, 4.1097, 3.7686, 1.8833, 3.7301, 3.7817, 3.7930, 3.5607], + device='cuda:1'), covar=tensor([0.0708, 0.0621, 0.1031, 0.4338, 0.0839, 0.0835, 0.1317, 0.0791], + device='cuda:1'), in_proj_covar=tensor([0.0475, 0.0389, 0.0400, 0.0491, 0.0393, 0.0393, 0.0385, 0.0342], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 14:11:52,789 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.385e+02 2.879e+02 3.637e+02 6.044e+02, threshold=5.759e+02, percent-clipped=0.0 +2023-02-06 14:11:58,855 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104067.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:12:18,829 INFO [train.py:901] (1/4) Epoch 13, batch 7100, loss[loss=0.2103, simple_loss=0.2949, pruned_loss=0.06288, over 8329.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3047, pruned_loss=0.07474, over 1612168.72 frames. ], batch size: 25, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:12:21,108 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104099.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:12:22,946 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104102.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:12:37,506 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104124.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 14:12:39,433 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104127.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:12:45,125 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8414, 1.7509, 2.5431, 1.7767, 1.3940, 2.5535, 0.6404, 1.4691], + device='cuda:1'), covar=tensor([0.2592, 0.1410, 0.0327, 0.1784, 0.3191, 0.0348, 0.2653, 0.1713], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0175, 0.0107, 0.0220, 0.0255, 0.0110, 0.0164, 0.0171], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 14:12:52,701 INFO [train.py:901] (1/4) Epoch 13, batch 7150, loss[loss=0.2367, simple_loss=0.3283, pruned_loss=0.07256, over 8342.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3048, pruned_loss=0.07474, over 1614169.87 frames. ], batch size: 26, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:13:00,089 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.541e+02 2.991e+02 4.071e+02 7.912e+02, threshold=5.982e+02, percent-clipped=4.0 +2023-02-06 14:13:16,456 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6012, 4.5992, 4.1487, 2.2819, 4.0964, 4.3742, 4.2217, 4.0912], + device='cuda:1'), covar=tensor([0.0681, 0.0482, 0.0920, 0.4085, 0.0770, 0.0851, 0.1154, 0.0742], + device='cuda:1'), in_proj_covar=tensor([0.0476, 0.0391, 0.0404, 0.0493, 0.0394, 0.0395, 0.0388, 0.0344], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 14:13:27,756 INFO [train.py:901] (1/4) Epoch 13, batch 7200, loss[loss=0.2185, simple_loss=0.303, pruned_loss=0.06699, over 8532.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3042, pruned_loss=0.07451, over 1615503.22 frames. ], batch size: 28, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:13:42,128 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104217.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:13:55,690 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104238.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:13:56,859 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 14:13:58,506 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104242.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:14:01,688 INFO [train.py:901] (1/4) Epoch 13, batch 7250, loss[loss=0.2751, simple_loss=0.3254, pruned_loss=0.1124, over 7120.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3052, pruned_loss=0.07555, over 1612730.58 frames. ], batch size: 71, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:14:09,625 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.469e+02 3.063e+02 3.939e+02 8.277e+02, threshold=6.126e+02, percent-clipped=7.0 +2023-02-06 14:14:37,316 INFO [train.py:901] (1/4) Epoch 13, batch 7300, loss[loss=0.2621, simple_loss=0.3364, pruned_loss=0.09392, over 8339.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.305, pruned_loss=0.07502, over 1610109.10 frames. ], batch size: 26, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:14:38,126 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1017, 1.6455, 1.6997, 1.5474, 0.8465, 1.5148, 1.7270, 1.5707], + device='cuda:1'), covar=tensor([0.0484, 0.1191, 0.1672, 0.1313, 0.0635, 0.1499, 0.0701, 0.0634], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0191, 0.0157, 0.0102, 0.0162, 0.0115, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 14:15:11,542 INFO [train.py:901] (1/4) Epoch 13, batch 7350, loss[loss=0.2003, simple_loss=0.2801, pruned_loss=0.06022, over 7649.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3036, pruned_loss=0.07391, over 1608934.93 frames. ], batch size: 19, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:15:14,979 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104352.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:15:15,793 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:15:19,099 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 2.439e+02 3.043e+02 3.823e+02 6.373e+02, threshold=6.086e+02, percent-clipped=2.0 +2023-02-06 14:15:19,892 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:15:33,234 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 14:15:46,158 INFO [train.py:901] (1/4) Epoch 13, batch 7400, loss[loss=0.2551, simple_loss=0.3373, pruned_loss=0.0864, over 8452.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3049, pruned_loss=0.07465, over 1611586.02 frames. ], batch size: 29, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:15:53,245 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 14:15:56,725 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104411.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:21,099 INFO [train.py:901] (1/4) Epoch 13, batch 7450, loss[loss=0.225, simple_loss=0.302, pruned_loss=0.07402, over 8333.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3036, pruned_loss=0.07424, over 1612570.36 frames. ], batch size: 25, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:16:29,260 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.494e+02 3.000e+02 3.814e+02 1.100e+03, threshold=5.999e+02, percent-clipped=4.0 +2023-02-06 14:16:33,244 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 14:16:35,445 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104467.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:39,565 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104473.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:40,155 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104474.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:48,678 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4991, 2.9039, 1.7808, 2.2095, 2.2294, 1.6217, 2.1082, 2.2376], + device='cuda:1'), covar=tensor([0.1330, 0.0248, 0.1060, 0.0649, 0.0711, 0.1267, 0.0861, 0.0911], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0237, 0.0322, 0.0297, 0.0301, 0.0322, 0.0342, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:16:55,884 INFO [train.py:901] (1/4) Epoch 13, batch 7500, loss[loss=0.214, simple_loss=0.2959, pruned_loss=0.06605, over 8329.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3023, pruned_loss=0.0739, over 1612790.22 frames. ], batch size: 25, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:16:56,750 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104498.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:56,770 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104498.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:57,438 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1862, 2.3188, 2.0109, 2.9944, 1.4132, 1.6270, 2.1054, 2.5404], + device='cuda:1'), covar=tensor([0.0715, 0.0934, 0.0946, 0.0312, 0.1198, 0.1484, 0.1028, 0.0749], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0209, 0.0253, 0.0210, 0.0215, 0.0253, 0.0256, 0.0216], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:17:14,590 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104523.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:17:16,624 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104526.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:17:25,703 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 +2023-02-06 14:17:28,999 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 14:17:30,657 INFO [train.py:901] (1/4) Epoch 13, batch 7550, loss[loss=0.2832, simple_loss=0.3465, pruned_loss=0.11, over 8477.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3023, pruned_loss=0.07419, over 1615460.61 frames. ], batch size: 25, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:17:37,882 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.482e+02 3.042e+02 4.105e+02 9.709e+02, threshold=6.085e+02, percent-clipped=7.0 +2023-02-06 14:18:05,094 INFO [train.py:901] (1/4) Epoch 13, batch 7600, loss[loss=0.2569, simple_loss=0.3299, pruned_loss=0.09198, over 8665.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3035, pruned_loss=0.07511, over 1615289.11 frames. ], batch size: 34, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:18:13,224 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104609.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:18:30,775 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:18:40,157 INFO [train.py:901] (1/4) Epoch 13, batch 7650, loss[loss=0.2851, simple_loss=0.3481, pruned_loss=0.111, over 8596.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3043, pruned_loss=0.07551, over 1612597.84 frames. ], batch size: 34, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:18:47,611 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 2.638e+02 3.280e+02 4.340e+02 1.130e+03, threshold=6.560e+02, percent-clipped=9.0 +2023-02-06 14:19:14,840 INFO [train.py:901] (1/4) Epoch 13, batch 7700, loss[loss=0.2249, simple_loss=0.3051, pruned_loss=0.07238, over 8460.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3041, pruned_loss=0.07482, over 1616675.52 frames. ], batch size: 25, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:19:33,199 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104723.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:19:37,759 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 14:19:37,981 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104730.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:19:49,822 INFO [train.py:901] (1/4) Epoch 13, batch 7750, loss[loss=0.2378, simple_loss=0.3207, pruned_loss=0.07747, over 8242.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3027, pruned_loss=0.0737, over 1617076.12 frames. ], batch size: 22, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:19:50,599 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104748.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:19:55,981 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:19:57,814 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.530e+02 2.944e+02 3.392e+02 9.198e+02, threshold=5.888e+02, percent-clipped=3.0 +2023-02-06 14:20:14,007 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104782.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:20:24,829 INFO [train.py:901] (1/4) Epoch 13, batch 7800, loss[loss=0.2052, simple_loss=0.2784, pruned_loss=0.06604, over 7193.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3016, pruned_loss=0.07272, over 1614752.14 frames. ], batch size: 16, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:20:31,892 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:20:57,827 INFO [train.py:901] (1/4) Epoch 13, batch 7850, loss[loss=0.2086, simple_loss=0.2982, pruned_loss=0.05951, over 8334.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3005, pruned_loss=0.07251, over 1612943.72 frames. ], batch size: 25, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:21:05,226 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.619e+02 3.074e+02 4.074e+02 1.012e+03, threshold=6.148e+02, percent-clipped=5.0 +2023-02-06 14:21:30,904 INFO [train.py:901] (1/4) Epoch 13, batch 7900, loss[loss=0.1958, simple_loss=0.2786, pruned_loss=0.05649, over 8242.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3008, pruned_loss=0.07241, over 1616187.71 frames. ], batch size: 22, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:21:46,719 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-06 14:22:04,224 INFO [train.py:901] (1/4) Epoch 13, batch 7950, loss[loss=0.2752, simple_loss=0.3392, pruned_loss=0.1056, over 8532.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3021, pruned_loss=0.07308, over 1617901.23 frames. ], batch size: 39, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:22:11,310 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.521e+02 3.027e+02 3.866e+02 6.555e+02, threshold=6.053e+02, percent-clipped=2.0 +2023-02-06 14:22:20,625 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6383, 1.4659, 4.8953, 1.7922, 4.3113, 4.0358, 4.3884, 4.2646], + device='cuda:1'), covar=tensor([0.0485, 0.4243, 0.0365, 0.3405, 0.0905, 0.0738, 0.0484, 0.0571], + device='cuda:1'), in_proj_covar=tensor([0.0522, 0.0586, 0.0609, 0.0553, 0.0634, 0.0539, 0.0527, 0.0592], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 14:22:37,769 INFO [train.py:901] (1/4) Epoch 13, batch 8000, loss[loss=0.2305, simple_loss=0.3078, pruned_loss=0.07664, over 8259.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3022, pruned_loss=0.0734, over 1618746.97 frames. ], batch size: 24, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:22:49,399 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 14:23:10,573 INFO [train.py:901] (1/4) Epoch 13, batch 8050, loss[loss=0.199, simple_loss=0.2772, pruned_loss=0.06035, over 7920.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.302, pruned_loss=0.07452, over 1604250.89 frames. ], batch size: 20, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:23:18,080 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.417e+02 2.946e+02 3.621e+02 6.025e+02, threshold=5.892e+02, percent-clipped=0.0 +2023-02-06 14:23:50,380 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 14:23:54,150 INFO [train.py:901] (1/4) Epoch 14, batch 0, loss[loss=0.233, simple_loss=0.3125, pruned_loss=0.07672, over 8514.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3125, pruned_loss=0.07672, over 8514.00 frames. ], batch size: 26, lr: 5.56e-03, grad_scale: 8.0 +2023-02-06 14:23:54,150 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 14:24:01,446 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3024, 1.4680, 2.2207, 1.0707, 1.6859, 1.5601, 1.3112, 1.6648], + device='cuda:1'), covar=tensor([0.1569, 0.2373, 0.0676, 0.3839, 0.1545, 0.2718, 0.1901, 0.1817], + device='cuda:1'), in_proj_covar=tensor([0.0499, 0.0545, 0.0539, 0.0600, 0.0623, 0.0566, 0.0493, 0.0619], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 14:24:05,201 INFO [train.py:935] (1/4) Epoch 14, validation: loss=0.184, simple_loss=0.2839, pruned_loss=0.04201, over 944034.00 frames. +2023-02-06 14:24:05,202 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 14:24:16,398 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-02-06 14:24:21,242 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 14:24:22,037 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([6.1102, 1.7713, 6.1070, 2.1963, 5.5557, 5.2153, 5.7232, 5.6111], + device='cuda:1'), covar=tensor([0.0383, 0.4288, 0.0313, 0.3153, 0.0825, 0.0698, 0.0378, 0.0417], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0580, 0.0608, 0.0551, 0.0632, 0.0536, 0.0523, 0.0589], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 14:24:38,586 INFO [train.py:901] (1/4) Epoch 14, batch 50, loss[loss=0.2062, simple_loss=0.274, pruned_loss=0.06918, over 7226.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.306, pruned_loss=0.07744, over 362828.66 frames. ], batch size: 16, lr: 5.56e-03, grad_scale: 8.0 +2023-02-06 14:24:54,214 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9785, 1.5214, 3.3754, 1.4167, 2.1604, 3.7332, 3.8094, 3.1415], + device='cuda:1'), covar=tensor([0.1094, 0.1640, 0.0326, 0.2056, 0.1071, 0.0219, 0.0440, 0.0592], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0304, 0.0265, 0.0295, 0.0281, 0.0241, 0.0362, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:24:54,762 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 14:24:58,151 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.784e+02 2.684e+02 3.092e+02 3.835e+02 7.852e+02, threshold=6.183e+02, percent-clipped=3.0 +2023-02-06 14:25:02,497 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6063, 1.9684, 3.2762, 1.3832, 2.2449, 2.0726, 1.7088, 2.2644], + device='cuda:1'), covar=tensor([0.1703, 0.2291, 0.0667, 0.4002, 0.1716, 0.2898, 0.1960, 0.2076], + device='cuda:1'), in_proj_covar=tensor([0.0498, 0.0541, 0.0535, 0.0597, 0.0623, 0.0565, 0.0490, 0.0615], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 14:25:14,421 INFO [train.py:901] (1/4) Epoch 14, batch 100, loss[loss=0.2409, simple_loss=0.3137, pruned_loss=0.08406, over 8588.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3072, pruned_loss=0.07637, over 645941.16 frames. ], batch size: 34, lr: 5.56e-03, grad_scale: 8.0 +2023-02-06 14:25:17,790 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 14:25:48,643 INFO [train.py:901] (1/4) Epoch 14, batch 150, loss[loss=0.1746, simple_loss=0.244, pruned_loss=0.0526, over 7432.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3048, pruned_loss=0.07569, over 855095.22 frames. ], batch size: 17, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:26:08,291 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.384e+02 2.990e+02 3.742e+02 5.781e+02, threshold=5.980e+02, percent-clipped=0.0 +2023-02-06 14:26:14,013 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-06 14:26:22,055 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-02-06 14:26:23,055 INFO [train.py:901] (1/4) Epoch 14, batch 200, loss[loss=0.1783, simple_loss=0.2587, pruned_loss=0.04898, over 8151.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3053, pruned_loss=0.07533, over 1029338.56 frames. ], batch size: 22, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:26:44,064 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 14:26:47,400 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1924, 1.3000, 1.5792, 1.2285, 0.7354, 1.3957, 1.2066, 0.9124], + device='cuda:1'), covar=tensor([0.0548, 0.1205, 0.1605, 0.1399, 0.0555, 0.1423, 0.0648, 0.0711], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0150, 0.0188, 0.0155, 0.0100, 0.0160, 0.0112, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 14:26:58,947 INFO [train.py:901] (1/4) Epoch 14, batch 250, loss[loss=0.2038, simple_loss=0.2886, pruned_loss=0.05952, over 7646.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3048, pruned_loss=0.07495, over 1156875.40 frames. ], batch size: 19, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:27:07,611 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 14:27:15,945 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 14:27:18,051 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.724e+02 2.546e+02 3.157e+02 4.204e+02 9.163e+02, threshold=6.313e+02, percent-clipped=6.0 +2023-02-06 14:27:31,870 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7758, 1.9105, 2.1584, 1.3780, 2.2536, 1.5188, 0.6931, 1.8973], + device='cuda:1'), covar=tensor([0.0406, 0.0260, 0.0168, 0.0397, 0.0255, 0.0647, 0.0626, 0.0209], + device='cuda:1'), in_proj_covar=tensor([0.0406, 0.0344, 0.0296, 0.0400, 0.0330, 0.0488, 0.0365, 0.0372], + device='cuda:1'), out_proj_covar=tensor([1.1341e-04, 9.3573e-05, 8.0427e-05, 1.0972e-04, 9.0763e-05, 1.4409e-04, + 1.0227e-04, 1.0270e-04], device='cuda:1') +2023-02-06 14:27:33,664 INFO [train.py:901] (1/4) Epoch 14, batch 300, loss[loss=0.2402, simple_loss=0.3197, pruned_loss=0.08037, over 8547.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3058, pruned_loss=0.0762, over 1255229.59 frames. ], batch size: 31, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:27:52,817 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105406.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:28:00,866 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7035, 1.7422, 1.5637, 1.9939, 1.3317, 1.4824, 1.6765, 1.9042], + device='cuda:1'), covar=tensor([0.0627, 0.0724, 0.0776, 0.0565, 0.0975, 0.1037, 0.0673, 0.0565], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0210, 0.0255, 0.0214, 0.0215, 0.0256, 0.0261, 0.0218], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:28:06,568 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 14:28:09,631 INFO [train.py:901] (1/4) Epoch 14, batch 350, loss[loss=0.2295, simple_loss=0.3132, pruned_loss=0.07289, over 8820.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.305, pruned_loss=0.07504, over 1337775.00 frames. ], batch size: 39, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:28:28,606 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 2.437e+02 2.818e+02 3.446e+02 5.751e+02, threshold=5.636e+02, percent-clipped=0.0 +2023-02-06 14:28:43,600 INFO [train.py:901] (1/4) Epoch 14, batch 400, loss[loss=0.2132, simple_loss=0.2739, pruned_loss=0.07625, over 7406.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3041, pruned_loss=0.07423, over 1398194.16 frames. ], batch size: 17, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:29:00,989 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:29:09,894 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5470, 2.7884, 1.9003, 2.1918, 2.2964, 1.5642, 2.0387, 2.1020], + device='cuda:1'), covar=tensor([0.1481, 0.0354, 0.1031, 0.0648, 0.0684, 0.1419, 0.0942, 0.1100], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0240, 0.0322, 0.0299, 0.0303, 0.0325, 0.0343, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:29:13,242 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:29:20,742 INFO [train.py:901] (1/4) Epoch 14, batch 450, loss[loss=0.1835, simple_loss=0.271, pruned_loss=0.04799, over 8238.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3033, pruned_loss=0.07376, over 1450177.27 frames. ], batch size: 22, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:29:40,049 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.497e+02 2.804e+02 3.770e+02 6.336e+02, threshold=5.609e+02, percent-clipped=1.0 +2023-02-06 14:29:55,212 INFO [train.py:901] (1/4) Epoch 14, batch 500, loss[loss=0.2206, simple_loss=0.2977, pruned_loss=0.07174, over 8615.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3034, pruned_loss=0.07367, over 1490785.12 frames. ], batch size: 34, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:30:29,387 INFO [train.py:901] (1/4) Epoch 14, batch 550, loss[loss=0.2523, simple_loss=0.3207, pruned_loss=0.09197, over 7990.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3031, pruned_loss=0.0743, over 1517702.56 frames. ], batch size: 21, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:30:50,299 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.442e+02 2.933e+02 3.700e+02 8.163e+02, threshold=5.867e+02, percent-clipped=3.0 +2023-02-06 14:31:05,192 INFO [train.py:901] (1/4) Epoch 14, batch 600, loss[loss=0.2202, simple_loss=0.2993, pruned_loss=0.07057, over 8239.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3029, pruned_loss=0.07371, over 1537580.09 frames. ], batch size: 22, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:31:18,452 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 14:31:39,813 INFO [train.py:901] (1/4) Epoch 14, batch 650, loss[loss=0.2408, simple_loss=0.3064, pruned_loss=0.08759, over 7650.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3021, pruned_loss=0.07287, over 1556194.93 frames. ], batch size: 19, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:31:54,386 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=105750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:32:01,339 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.402e+02 3.000e+02 3.711e+02 7.109e+02, threshold=6.000e+02, percent-clipped=4.0 +2023-02-06 14:32:17,053 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.05 vs. limit=5.0 +2023-02-06 14:32:17,354 INFO [train.py:901] (1/4) Epoch 14, batch 700, loss[loss=0.2255, simple_loss=0.3048, pruned_loss=0.07306, over 8483.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.302, pruned_loss=0.0727, over 1571424.09 frames. ], batch size: 26, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:32:37,865 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:32:42,735 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105817.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:32:51,384 INFO [train.py:901] (1/4) Epoch 14, batch 750, loss[loss=0.2318, simple_loss=0.3071, pruned_loss=0.07819, over 8087.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3019, pruned_loss=0.07331, over 1583720.85 frames. ], batch size: 21, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:32:51,591 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4877, 2.8506, 1.8330, 2.1555, 2.1165, 1.5774, 1.8796, 2.3504], + device='cuda:1'), covar=tensor([0.1694, 0.0484, 0.1318, 0.0830, 0.0913, 0.1657, 0.1453, 0.0981], + device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0236, 0.0318, 0.0295, 0.0299, 0.0320, 0.0337, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:33:03,871 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=105848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:33:06,441 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 14:33:11,302 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.389e+02 2.459e+02 2.898e+02 3.725e+02 7.154e+02, threshold=5.796e+02, percent-clipped=4.0 +2023-02-06 14:33:15,486 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=105864.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:33:16,040 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 14:33:16,238 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=105865.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:33:27,181 INFO [train.py:901] (1/4) Epoch 14, batch 800, loss[loss=0.223, simple_loss=0.3105, pruned_loss=0.0678, over 8467.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3016, pruned_loss=0.07299, over 1590389.87 frames. ], batch size: 29, lr: 5.54e-03, grad_scale: 16.0 +2023-02-06 14:34:02,180 INFO [train.py:901] (1/4) Epoch 14, batch 850, loss[loss=0.1704, simple_loss=0.2481, pruned_loss=0.04633, over 7976.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3005, pruned_loss=0.07249, over 1594180.84 frames. ], batch size: 21, lr: 5.54e-03, grad_scale: 16.0 +2023-02-06 14:34:20,964 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.803e+02 2.477e+02 2.961e+02 4.061e+02 6.411e+02, threshold=5.921e+02, percent-clipped=4.0 +2023-02-06 14:34:24,609 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=105963.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:34:36,585 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=105979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:34:37,077 INFO [train.py:901] (1/4) Epoch 14, batch 900, loss[loss=0.1814, simple_loss=0.2539, pruned_loss=0.05446, over 7538.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3004, pruned_loss=0.07236, over 1596793.22 frames. ], batch size: 18, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:35:03,393 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1808, 1.0435, 1.2460, 1.1029, 0.9742, 1.3143, 0.0262, 0.9547], + device='cuda:1'), covar=tensor([0.2172, 0.1594, 0.0620, 0.1048, 0.3256, 0.0569, 0.2852, 0.1615], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0177, 0.0108, 0.0220, 0.0256, 0.0111, 0.0164, 0.0174], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 14:35:14,900 INFO [train.py:901] (1/4) Epoch 14, batch 950, loss[loss=0.2176, simple_loss=0.3019, pruned_loss=0.06666, over 8190.00 frames. ], tot_loss[loss=0.224, simple_loss=0.302, pruned_loss=0.07303, over 1604727.31 frames. ], batch size: 23, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:35:33,973 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.617e+02 3.202e+02 4.119e+02 6.844e+02, threshold=6.403e+02, percent-clipped=3.0 +2023-02-06 14:35:38,929 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 14:35:49,306 INFO [train.py:901] (1/4) Epoch 14, batch 1000, loss[loss=0.2385, simple_loss=0.3137, pruned_loss=0.08164, over 7983.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3018, pruned_loss=0.07319, over 1608731.87 frames. ], batch size: 21, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:36:00,603 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106095.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:36:14,292 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 14:36:20,062 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106121.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:36:26,961 INFO [train.py:901] (1/4) Epoch 14, batch 1050, loss[loss=0.1939, simple_loss=0.262, pruned_loss=0.06286, over 7707.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3017, pruned_loss=0.07305, over 1608683.58 frames. ], batch size: 18, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:36:26,975 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 14:36:37,960 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106146.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:36:43,490 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106154.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:36:46,240 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.417e+02 2.951e+02 3.593e+02 9.096e+02, threshold=5.903e+02, percent-clipped=2.0 +2023-02-06 14:36:48,433 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:37:01,608 INFO [train.py:901] (1/4) Epoch 14, batch 1100, loss[loss=0.1858, simple_loss=0.265, pruned_loss=0.05331, over 7800.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3013, pruned_loss=0.07288, over 1611066.47 frames. ], batch size: 19, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:37:29,717 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106219.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:37:35,896 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 14:37:38,713 INFO [train.py:901] (1/4) Epoch 14, batch 1150, loss[loss=0.1847, simple_loss=0.2702, pruned_loss=0.0496, over 8031.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3005, pruned_loss=0.07232, over 1609560.37 frames. ], batch size: 22, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:37:42,400 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106235.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:37:46,538 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-02-06 14:37:49,104 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106244.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:37:58,401 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.475e+02 3.133e+02 3.919e+02 6.906e+02, threshold=6.266e+02, percent-clipped=3.0 +2023-02-06 14:38:00,001 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106260.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:38:06,193 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106269.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:38:10,868 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106276.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:38:13,353 INFO [train.py:901] (1/4) Epoch 14, batch 1200, loss[loss=0.2331, simple_loss=0.3119, pruned_loss=0.07715, over 8105.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.301, pruned_loss=0.0729, over 1611484.92 frames. ], batch size: 23, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:38:17,553 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:38:42,805 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3012, 1.5717, 4.1594, 1.7866, 2.3592, 4.7226, 4.6881, 4.0111], + device='cuda:1'), covar=tensor([0.1011, 0.1769, 0.0303, 0.1992, 0.1209, 0.0185, 0.0398, 0.0576], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0307, 0.0267, 0.0296, 0.0284, 0.0245, 0.0365, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:38:47,948 INFO [train.py:901] (1/4) Epoch 14, batch 1250, loss[loss=0.2498, simple_loss=0.3159, pruned_loss=0.09185, over 7309.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3028, pruned_loss=0.0739, over 1614631.26 frames. ], batch size: 71, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:39:05,959 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106354.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:39:08,474 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.539e+02 3.303e+02 4.386e+02 1.450e+03, threshold=6.607e+02, percent-clipped=4.0 +2023-02-06 14:39:24,633 INFO [train.py:901] (1/4) Epoch 14, batch 1300, loss[loss=0.2278, simple_loss=0.3077, pruned_loss=0.07393, over 8569.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3022, pruned_loss=0.07362, over 1615309.55 frames. ], batch size: 31, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:39:55,727 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6775, 1.9104, 2.0812, 1.3381, 2.1841, 1.5056, 0.5367, 1.8136], + device='cuda:1'), covar=tensor([0.0437, 0.0230, 0.0206, 0.0373, 0.0242, 0.0600, 0.0579, 0.0183], + device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0346, 0.0302, 0.0404, 0.0337, 0.0491, 0.0366, 0.0376], + device='cuda:1'), out_proj_covar=tensor([1.1432e-04, 9.3765e-05, 8.2173e-05, 1.1042e-04, 9.2543e-05, 1.4467e-04, + 1.0230e-04, 1.0355e-04], device='cuda:1') +2023-02-06 14:39:58,990 INFO [train.py:901] (1/4) Epoch 14, batch 1350, loss[loss=0.2198, simple_loss=0.3069, pruned_loss=0.06638, over 8466.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3015, pruned_loss=0.07289, over 1618157.73 frames. ], batch size: 25, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:40:05,432 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106439.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:40:19,201 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.550e+02 3.060e+02 3.665e+02 8.767e+02, threshold=6.121e+02, percent-clipped=1.0 +2023-02-06 14:40:29,773 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:40:35,138 INFO [train.py:901] (1/4) Epoch 14, batch 1400, loss[loss=0.1916, simple_loss=0.2637, pruned_loss=0.0598, over 7553.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3016, pruned_loss=0.07303, over 1615121.44 frames. ], batch size: 18, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:41:05,456 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 14:41:07,391 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106525.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:10,539 INFO [train.py:901] (1/4) Epoch 14, batch 1450, loss[loss=0.2697, simple_loss=0.344, pruned_loss=0.0977, over 8538.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3024, pruned_loss=0.0734, over 1618269.21 frames. ], batch size: 39, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:41:11,252 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 14:41:12,179 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:24,668 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106550.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:27,429 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106554.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:29,526 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106557.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:29,959 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.831e+02 2.546e+02 3.123e+02 4.151e+02 8.254e+02, threshold=6.246e+02, percent-clipped=6.0 +2023-02-06 14:41:47,568 INFO [train.py:901] (1/4) Epoch 14, batch 1500, loss[loss=0.256, simple_loss=0.3275, pruned_loss=0.09221, over 8717.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3025, pruned_loss=0.07391, over 1613805.52 frames. ], batch size: 34, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:42:22,567 INFO [train.py:901] (1/4) Epoch 14, batch 1550, loss[loss=0.2327, simple_loss=0.3215, pruned_loss=0.07194, over 8545.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3029, pruned_loss=0.07407, over 1618927.98 frames. ], batch size: 31, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:42:22,643 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:42:41,337 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.593e+02 3.196e+02 4.114e+02 8.054e+02, threshold=6.391e+02, percent-clipped=4.0 +2023-02-06 14:42:52,894 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9790, 1.5144, 3.5327, 1.6430, 2.3846, 3.7970, 3.8848, 3.3130], + device='cuda:1'), covar=tensor([0.1027, 0.1595, 0.0272, 0.1842, 0.0956, 0.0211, 0.0496, 0.0545], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0303, 0.0263, 0.0293, 0.0281, 0.0242, 0.0362, 0.0293], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 14:42:56,719 INFO [train.py:901] (1/4) Epoch 14, batch 1600, loss[loss=0.2913, simple_loss=0.355, pruned_loss=0.1138, over 8436.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3037, pruned_loss=0.07441, over 1617879.39 frames. ], batch size: 39, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:43:10,336 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:43:32,335 INFO [train.py:901] (1/4) Epoch 14, batch 1650, loss[loss=0.2055, simple_loss=0.2813, pruned_loss=0.06485, over 7977.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3031, pruned_loss=0.07419, over 1615698.57 frames. ], batch size: 21, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:43:35,237 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7700, 1.7800, 2.1008, 1.8024, 1.0537, 1.7995, 2.3832, 2.1605], + device='cuda:1'), covar=tensor([0.0430, 0.1176, 0.1582, 0.1242, 0.0567, 0.1425, 0.0557, 0.0556], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0157, 0.0101, 0.0162, 0.0113, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 14:43:42,577 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:43:51,909 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.493e+02 3.038e+02 4.078e+02 1.080e+03, threshold=6.076e+02, percent-clipped=3.0 +2023-02-06 14:43:52,351 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.62 vs. limit=5.0 +2023-02-06 14:43:53,535 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6352, 1.8921, 2.0627, 1.2631, 2.1401, 1.4325, 0.5292, 1.8428], + device='cuda:1'), covar=tensor([0.0417, 0.0283, 0.0228, 0.0399, 0.0291, 0.0685, 0.0595, 0.0207], + device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0349, 0.0304, 0.0406, 0.0338, 0.0492, 0.0370, 0.0375], + device='cuda:1'), out_proj_covar=tensor([1.1478e-04, 9.4922e-05, 8.2619e-05, 1.1079e-04, 9.2894e-05, 1.4505e-04, + 1.0328e-04, 1.0334e-04], device='cuda:1') +2023-02-06 14:44:06,421 INFO [train.py:901] (1/4) Epoch 14, batch 1700, loss[loss=0.2273, simple_loss=0.3164, pruned_loss=0.06915, over 8495.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3037, pruned_loss=0.07396, over 1617309.03 frames. ], batch size: 28, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:44:28,260 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:44:31,567 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:44:33,567 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106816.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:44:43,433 INFO [train.py:901] (1/4) Epoch 14, batch 1750, loss[loss=0.2281, simple_loss=0.3177, pruned_loss=0.06931, over 8258.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3029, pruned_loss=0.07338, over 1618845.08 frames. ], batch size: 24, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:44:47,853 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106835.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:45:04,128 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.358e+02 2.865e+02 3.554e+02 7.426e+02, threshold=5.730e+02, percent-clipped=3.0 +2023-02-06 14:45:18,432 INFO [train.py:901] (1/4) Epoch 14, batch 1800, loss[loss=0.1794, simple_loss=0.2613, pruned_loss=0.04872, over 7927.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3002, pruned_loss=0.07226, over 1611413.07 frames. ], batch size: 20, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:45:54,583 INFO [train.py:901] (1/4) Epoch 14, batch 1850, loss[loss=0.2275, simple_loss=0.3088, pruned_loss=0.07313, over 8501.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3016, pruned_loss=0.07281, over 1615881.85 frames. ], batch size: 26, lr: 5.51e-03, grad_scale: 4.0 +2023-02-06 14:45:55,505 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106931.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:46:16,023 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 2.570e+02 3.068e+02 3.847e+02 1.325e+03, threshold=6.136e+02, percent-clipped=4.0 +2023-02-06 14:46:29,529 INFO [train.py:901] (1/4) Epoch 14, batch 1900, loss[loss=0.1959, simple_loss=0.2731, pruned_loss=0.05932, over 7913.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3037, pruned_loss=0.07428, over 1618785.22 frames. ], batch size: 20, lr: 5.51e-03, grad_scale: 4.0 +2023-02-06 14:46:43,885 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107001.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:46:47,082 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 14:46:59,783 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 14:47:01,262 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107026.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:47:03,727 INFO [train.py:901] (1/4) Epoch 14, batch 1950, loss[loss=0.2258, simple_loss=0.3111, pruned_loss=0.07029, over 8327.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3031, pruned_loss=0.0735, over 1620337.14 frames. ], batch size: 25, lr: 5.51e-03, grad_scale: 4.0 +2023-02-06 14:47:19,881 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 14:47:26,063 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.386e+02 2.840e+02 3.483e+02 6.138e+02, threshold=5.681e+02, percent-clipped=1.0 +2023-02-06 14:47:31,983 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107069.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:47:39,105 INFO [train.py:901] (1/4) Epoch 14, batch 2000, loss[loss=0.2224, simple_loss=0.3005, pruned_loss=0.0721, over 7655.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3037, pruned_loss=0.07395, over 1618692.12 frames. ], batch size: 19, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:47:48,656 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:48:12,599 INFO [train.py:901] (1/4) Epoch 14, batch 2050, loss[loss=0.258, simple_loss=0.3297, pruned_loss=0.09314, over 8526.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3036, pruned_loss=0.07372, over 1618907.16 frames. ], batch size: 28, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:48:19,570 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4131, 1.5864, 1.3642, 1.8973, 0.8550, 1.2622, 1.2452, 1.5221], + device='cuda:1'), covar=tensor([0.0801, 0.0701, 0.1019, 0.0480, 0.1106, 0.1356, 0.0794, 0.0734], + device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0203, 0.0245, 0.0207, 0.0211, 0.0247, 0.0251, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:48:23,060 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6422, 1.6218, 2.8064, 1.1638, 2.1518, 2.9683, 3.1137, 2.5039], + device='cuda:1'), covar=tensor([0.1206, 0.1403, 0.0408, 0.2253, 0.0926, 0.0323, 0.0575, 0.0687], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0304, 0.0264, 0.0292, 0.0279, 0.0241, 0.0362, 0.0293], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 14:48:26,611 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9465, 1.6271, 2.1053, 1.8192, 2.0072, 1.9300, 1.6725, 0.7742], + device='cuda:1'), covar=tensor([0.4743, 0.4067, 0.1530, 0.2809, 0.2079, 0.2690, 0.1930, 0.4385], + device='cuda:1'), in_proj_covar=tensor([0.0897, 0.0898, 0.0746, 0.0868, 0.0956, 0.0825, 0.0710, 0.0779], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 14:48:27,933 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1975, 2.4022, 2.0435, 2.9950, 1.3839, 1.8232, 2.0246, 2.5735], + device='cuda:1'), covar=tensor([0.0637, 0.0869, 0.0804, 0.0334, 0.1148, 0.1180, 0.0995, 0.0761], + device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0202, 0.0245, 0.0207, 0.0210, 0.0247, 0.0251, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:48:32,521 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107158.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 14:48:34,367 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.411e+02 3.055e+02 3.713e+02 7.642e+02, threshold=6.109e+02, percent-clipped=4.0 +2023-02-06 14:48:42,096 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107170.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:48:49,408 INFO [train.py:901] (1/4) Epoch 14, batch 2100, loss[loss=0.3015, simple_loss=0.3716, pruned_loss=0.1157, over 8640.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3029, pruned_loss=0.07305, over 1621329.06 frames. ], batch size: 39, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:48:54,367 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107187.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:48:57,046 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6730, 2.2614, 4.1515, 1.4528, 2.9556, 2.2585, 1.8849, 2.7914], + device='cuda:1'), covar=tensor([0.1737, 0.2141, 0.0694, 0.3808, 0.1547, 0.2727, 0.1821, 0.2204], + device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0542, 0.0535, 0.0592, 0.0618, 0.0559, 0.0487, 0.0614], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 14:49:00,233 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6389, 1.4330, 2.8568, 1.3460, 2.0695, 3.0888, 3.1448, 2.6019], + device='cuda:1'), covar=tensor([0.1125, 0.1448, 0.0418, 0.2005, 0.0933, 0.0288, 0.0672, 0.0628], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0305, 0.0266, 0.0293, 0.0281, 0.0242, 0.0363, 0.0293], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 14:49:11,188 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107212.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:49:23,204 INFO [train.py:901] (1/4) Epoch 14, batch 2150, loss[loss=0.2458, simple_loss=0.3286, pruned_loss=0.08152, over 8495.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3016, pruned_loss=0.07242, over 1622343.86 frames. ], batch size: 28, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:49:44,434 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 2.615e+02 3.041e+02 3.823e+02 8.460e+02, threshold=6.081e+02, percent-clipped=1.0 +2023-02-06 14:49:58,912 INFO [train.py:901] (1/4) Epoch 14, batch 2200, loss[loss=0.208, simple_loss=0.2883, pruned_loss=0.0638, over 7978.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.302, pruned_loss=0.07317, over 1616804.10 frames. ], batch size: 21, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:50:23,273 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6796, 1.9768, 2.1500, 1.2760, 2.3174, 1.4604, 0.7046, 1.9029], + device='cuda:1'), covar=tensor([0.0483, 0.0283, 0.0194, 0.0436, 0.0293, 0.0719, 0.0672, 0.0251], + device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0351, 0.0306, 0.0407, 0.0337, 0.0496, 0.0372, 0.0378], + device='cuda:1'), out_proj_covar=tensor([1.1456e-04, 9.5359e-05, 8.3165e-05, 1.1110e-04, 9.2411e-05, 1.4607e-04, + 1.0394e-04, 1.0421e-04], device='cuda:1') +2023-02-06 14:50:34,524 INFO [train.py:901] (1/4) Epoch 14, batch 2250, loss[loss=0.1862, simple_loss=0.2617, pruned_loss=0.0554, over 7437.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3015, pruned_loss=0.07305, over 1616920.95 frames. ], batch size: 17, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:50:54,555 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.599e+02 3.319e+02 4.071e+02 1.027e+03, threshold=6.637e+02, percent-clipped=7.0 +2023-02-06 14:51:08,891 INFO [train.py:901] (1/4) Epoch 14, batch 2300, loss[loss=0.1558, simple_loss=0.2393, pruned_loss=0.03619, over 7932.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3005, pruned_loss=0.07215, over 1613857.15 frames. ], batch size: 20, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:51:44,729 INFO [train.py:901] (1/4) Epoch 14, batch 2350, loss[loss=0.2724, simple_loss=0.3418, pruned_loss=0.1015, over 8644.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3015, pruned_loss=0.07256, over 1618860.24 frames. ], batch size: 34, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:52:04,938 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.358e+02 2.889e+02 3.449e+02 7.134e+02, threshold=5.779e+02, percent-clipped=1.0 +2023-02-06 14:52:15,943 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8248, 5.8760, 5.1069, 2.3958, 5.1716, 5.5514, 5.4191, 5.2453], + device='cuda:1'), covar=tensor([0.0484, 0.0360, 0.0875, 0.4218, 0.0636, 0.0638, 0.0931, 0.0670], + device='cuda:1'), in_proj_covar=tensor([0.0483, 0.0400, 0.0404, 0.0496, 0.0398, 0.0400, 0.0384, 0.0345], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 14:52:18,380 INFO [train.py:901] (1/4) Epoch 14, batch 2400, loss[loss=0.273, simple_loss=0.3262, pruned_loss=0.1099, over 7961.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3017, pruned_loss=0.07302, over 1617841.12 frames. ], batch size: 21, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:52:34,429 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107502.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:52:43,465 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107514.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:52:53,760 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107528.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:52:54,997 INFO [train.py:901] (1/4) Epoch 14, batch 2450, loss[loss=0.2141, simple_loss=0.3034, pruned_loss=0.06238, over 8529.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3032, pruned_loss=0.07431, over 1617039.67 frames. ], batch size: 28, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:53:16,529 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.477e+02 3.089e+02 4.011e+02 1.178e+03, threshold=6.179e+02, percent-clipped=8.0 +2023-02-06 14:53:29,785 INFO [train.py:901] (1/4) Epoch 14, batch 2500, loss[loss=0.2196, simple_loss=0.3011, pruned_loss=0.06901, over 8732.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3041, pruned_loss=0.0749, over 1616099.90 frames. ], batch size: 34, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:53:55,582 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107617.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:54:03,476 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107629.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:54:03,967 INFO [train.py:901] (1/4) Epoch 14, batch 2550, loss[loss=0.2123, simple_loss=0.2978, pruned_loss=0.06339, over 8469.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3027, pruned_loss=0.07374, over 1616498.41 frames. ], batch size: 25, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:54:26,274 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.642e+02 3.253e+02 4.518e+02 1.030e+03, threshold=6.506e+02, percent-clipped=5.0 +2023-02-06 14:54:31,308 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2024, 2.4345, 2.0125, 2.9658, 1.2626, 1.7012, 2.0175, 2.4697], + device='cuda:1'), covar=tensor([0.0634, 0.0734, 0.0921, 0.0344, 0.1195, 0.1337, 0.0881, 0.0833], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0206, 0.0251, 0.0212, 0.0216, 0.0253, 0.0255, 0.0216], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:54:37,782 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107677.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:54:39,570 INFO [train.py:901] (1/4) Epoch 14, batch 2600, loss[loss=0.2275, simple_loss=0.2894, pruned_loss=0.08279, over 7667.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3021, pruned_loss=0.0738, over 1614418.39 frames. ], batch size: 19, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:54:48,870 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-02-06 14:55:12,844 INFO [train.py:901] (1/4) Epoch 14, batch 2650, loss[loss=0.2066, simple_loss=0.2744, pruned_loss=0.06937, over 7426.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3021, pruned_loss=0.07404, over 1610517.36 frames. ], batch size: 17, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:55:30,857 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:55:34,866 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.443e+02 2.980e+02 3.881e+02 9.981e+02, threshold=5.960e+02, percent-clipped=6.0 +2023-02-06 14:55:38,604 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5531, 1.6896, 1.7598, 1.2102, 1.8200, 1.4130, 0.8446, 1.6811], + device='cuda:1'), covar=tensor([0.0355, 0.0219, 0.0160, 0.0345, 0.0244, 0.0479, 0.0508, 0.0195], + device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0353, 0.0308, 0.0411, 0.0341, 0.0499, 0.0376, 0.0381], + device='cuda:1'), out_proj_covar=tensor([1.1548e-04, 9.5921e-05, 8.3857e-05, 1.1200e-04, 9.3445e-05, 1.4695e-04, + 1.0487e-04, 1.0510e-04], device='cuda:1') +2023-02-06 14:55:49,934 INFO [train.py:901] (1/4) Epoch 14, batch 2700, loss[loss=0.243, simple_loss=0.3145, pruned_loss=0.08568, over 8245.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3011, pruned_loss=0.07372, over 1607385.21 frames. ], batch size: 22, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:56:04,448 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0595, 4.1590, 2.8069, 3.0046, 3.1407, 2.2656, 2.7357, 3.1471], + device='cuda:1'), covar=tensor([0.1467, 0.0258, 0.0751, 0.0653, 0.0642, 0.1120, 0.1005, 0.0784], + device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0230, 0.0316, 0.0295, 0.0296, 0.0319, 0.0336, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:56:23,702 INFO [train.py:901] (1/4) Epoch 14, batch 2750, loss[loss=0.1977, simple_loss=0.2853, pruned_loss=0.05507, over 7658.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3029, pruned_loss=0.07383, over 1616655.40 frames. ], batch size: 19, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:56:44,693 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.404e+02 2.918e+02 3.592e+02 1.217e+03, threshold=5.837e+02, percent-clipped=4.0 +2023-02-06 14:56:53,310 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107872.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:56:54,154 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107873.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:56:59,460 INFO [train.py:901] (1/4) Epoch 14, batch 2800, loss[loss=0.2378, simple_loss=0.315, pruned_loss=0.08026, over 8669.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3031, pruned_loss=0.07419, over 1613947.50 frames. ], batch size: 39, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:57:03,900 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107885.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:57:12,673 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107898.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:57:20,767 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107910.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:57:33,811 INFO [train.py:901] (1/4) Epoch 14, batch 2850, loss[loss=0.2112, simple_loss=0.2947, pruned_loss=0.06391, over 8248.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3025, pruned_loss=0.07367, over 1610500.08 frames. ], batch size: 22, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 14:57:54,109 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.478e+02 3.087e+02 3.919e+02 8.173e+02, threshold=6.173e+02, percent-clipped=5.0 +2023-02-06 14:58:01,786 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2031, 2.2694, 1.7220, 2.0389, 1.8446, 1.3740, 1.6160, 1.6933], + device='cuda:1'), covar=tensor([0.1126, 0.0316, 0.0958, 0.0462, 0.0641, 0.1280, 0.0927, 0.0787], + device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0227, 0.0315, 0.0293, 0.0293, 0.0317, 0.0337, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 14:58:03,816 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1391, 2.4503, 3.6761, 1.9058, 2.8600, 2.4418, 2.2410, 2.6405], + device='cuda:1'), covar=tensor([0.1138, 0.1585, 0.0503, 0.2487, 0.1189, 0.1919, 0.1229, 0.1793], + device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0543, 0.0535, 0.0589, 0.0619, 0.0555, 0.0486, 0.0612], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 14:58:05,775 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9350, 2.2156, 1.8087, 2.7875, 1.3495, 1.5994, 1.9499, 2.2151], + device='cuda:1'), covar=tensor([0.0715, 0.0858, 0.0978, 0.0387, 0.1200, 0.1410, 0.0908, 0.0813], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0204, 0.0248, 0.0210, 0.0212, 0.0248, 0.0252, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 14:58:08,144 INFO [train.py:901] (1/4) Epoch 14, batch 2900, loss[loss=0.2353, simple_loss=0.3135, pruned_loss=0.07852, over 8353.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3017, pruned_loss=0.07347, over 1607902.35 frames. ], batch size: 26, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 14:58:12,782 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107987.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:58:27,891 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 14:58:38,640 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108021.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:58:44,583 INFO [train.py:901] (1/4) Epoch 14, batch 2950, loss[loss=0.2462, simple_loss=0.3141, pruned_loss=0.08918, over 8205.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3008, pruned_loss=0.07344, over 1602102.65 frames. ], batch size: 23, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 14:59:04,840 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.696e+02 3.199e+02 4.019e+02 8.231e+02, threshold=6.398e+02, percent-clipped=3.0 +2023-02-06 14:59:18,166 INFO [train.py:901] (1/4) Epoch 14, batch 3000, loss[loss=0.2274, simple_loss=0.3175, pruned_loss=0.0686, over 8491.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3001, pruned_loss=0.07275, over 1603511.63 frames. ], batch size: 25, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 14:59:18,166 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 14:59:30,508 INFO [train.py:935] (1/4) Epoch 14, validation: loss=0.1827, simple_loss=0.283, pruned_loss=0.04121, over 944034.00 frames. +2023-02-06 14:59:30,510 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 14:59:43,701 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108099.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:00:05,762 INFO [train.py:901] (1/4) Epoch 14, batch 3050, loss[loss=0.2368, simple_loss=0.3124, pruned_loss=0.0806, over 8396.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3006, pruned_loss=0.07323, over 1603251.80 frames. ], batch size: 49, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 15:00:10,186 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.42 vs. limit=5.0 +2023-02-06 15:00:10,682 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108136.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:00:28,076 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.670e+02 3.118e+02 3.835e+02 7.160e+02, threshold=6.236e+02, percent-clipped=1.0 +2023-02-06 15:00:41,679 INFO [train.py:901] (1/4) Epoch 14, batch 3100, loss[loss=0.1927, simple_loss=0.2604, pruned_loss=0.06247, over 7697.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.301, pruned_loss=0.07332, over 1605743.87 frames. ], batch size: 18, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 15:00:49,460 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-02-06 15:01:04,702 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:01:15,963 INFO [train.py:901] (1/4) Epoch 14, batch 3150, loss[loss=0.249, simple_loss=0.3352, pruned_loss=0.08136, over 8198.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3014, pruned_loss=0.07348, over 1604923.79 frames. ], batch size: 23, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 15:01:24,679 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108243.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:01:37,167 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.570e+02 3.163e+02 4.155e+02 7.848e+02, threshold=6.326e+02, percent-clipped=5.0 +2023-02-06 15:01:43,492 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108268.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:01:46,379 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 15:01:51,511 INFO [train.py:901] (1/4) Epoch 14, batch 3200, loss[loss=0.2061, simple_loss=0.2877, pruned_loss=0.06222, over 8719.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.302, pruned_loss=0.07368, over 1609556.13 frames. ], batch size: 30, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 15:01:51,735 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5183, 1.4672, 1.7656, 1.3076, 1.1421, 1.7859, 0.1278, 1.0647], + device='cuda:1'), covar=tensor([0.2445, 0.1884, 0.0587, 0.1457, 0.3987, 0.0564, 0.2918, 0.1735], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0177, 0.0109, 0.0220, 0.0261, 0.0111, 0.0163, 0.0174], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 15:02:06,043 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5501, 2.2281, 3.3905, 2.5985, 2.9496, 2.3912, 1.9723, 1.9802], + device='cuda:1'), covar=tensor([0.3969, 0.4113, 0.1366, 0.2714, 0.2131, 0.2181, 0.1627, 0.4078], + device='cuda:1'), in_proj_covar=tensor([0.0902, 0.0901, 0.0740, 0.0868, 0.0957, 0.0831, 0.0712, 0.0778], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 15:02:25,763 INFO [train.py:901] (1/4) Epoch 14, batch 3250, loss[loss=0.1968, simple_loss=0.2778, pruned_loss=0.05792, over 8459.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3033, pruned_loss=0.07428, over 1608973.05 frames. ], batch size: 27, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:02:35,708 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:02:47,022 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 2.638e+02 3.239e+02 4.086e+02 1.012e+03, threshold=6.478e+02, percent-clipped=4.0 +2023-02-06 15:03:02,204 INFO [train.py:901] (1/4) Epoch 14, batch 3300, loss[loss=0.2353, simple_loss=0.3221, pruned_loss=0.07427, over 8030.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3028, pruned_loss=0.07363, over 1612470.19 frames. ], batch size: 22, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:03:11,260 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:03:27,903 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:03:36,466 INFO [train.py:901] (1/4) Epoch 14, batch 3350, loss[loss=0.1959, simple_loss=0.2811, pruned_loss=0.05538, over 8127.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3027, pruned_loss=0.07336, over 1616313.45 frames. ], batch size: 22, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:03:49,808 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108450.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:03:57,188 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.656e+02 3.299e+02 4.467e+02 8.781e+02, threshold=6.597e+02, percent-clipped=5.0 +2023-02-06 15:04:04,263 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108470.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:04:10,796 INFO [train.py:901] (1/4) Epoch 14, batch 3400, loss[loss=0.2212, simple_loss=0.3129, pruned_loss=0.06471, over 8328.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3041, pruned_loss=0.07418, over 1618322.82 frames. ], batch size: 25, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:04:22,161 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108495.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:04:46,713 INFO [train.py:901] (1/4) Epoch 14, batch 3450, loss[loss=0.2715, simple_loss=0.326, pruned_loss=0.1085, over 7055.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3036, pruned_loss=0.07405, over 1614803.28 frames. ], batch size: 71, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:05:07,901 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.481e+02 3.055e+02 3.627e+02 7.933e+02, threshold=6.110e+02, percent-clipped=3.0 +2023-02-06 15:05:21,984 INFO [train.py:901] (1/4) Epoch 14, batch 3500, loss[loss=0.2831, simple_loss=0.3556, pruned_loss=0.1053, over 8460.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3042, pruned_loss=0.07438, over 1614618.42 frames. ], batch size: 27, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:05:29,151 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 15:05:31,989 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108595.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:05:57,729 INFO [train.py:901] (1/4) Epoch 14, batch 3550, loss[loss=0.2186, simple_loss=0.3063, pruned_loss=0.06547, over 8334.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3034, pruned_loss=0.07417, over 1615535.75 frames. ], batch size: 25, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:06:17,929 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.417e+02 3.151e+02 4.175e+02 8.210e+02, threshold=6.301e+02, percent-clipped=3.0 +2023-02-06 15:06:28,318 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2092, 1.3526, 3.3459, 1.0399, 2.9314, 2.7743, 3.0131, 2.9510], + device='cuda:1'), covar=tensor([0.0801, 0.4080, 0.0858, 0.4044, 0.1504, 0.1233, 0.0765, 0.0880], + device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0598, 0.0614, 0.0560, 0.0636, 0.0545, 0.0534, 0.0598], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 15:06:31,420 INFO [train.py:901] (1/4) Epoch 14, batch 3600, loss[loss=0.1942, simple_loss=0.2688, pruned_loss=0.0598, over 7800.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.303, pruned_loss=0.07386, over 1615770.46 frames. ], batch size: 20, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:06:36,125 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108687.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:07:07,041 INFO [train.py:901] (1/4) Epoch 14, batch 3650, loss[loss=0.2226, simple_loss=0.3034, pruned_loss=0.07089, over 8458.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3019, pruned_loss=0.07324, over 1612950.93 frames. ], batch size: 27, lr: 5.46e-03, grad_scale: 8.0 +2023-02-06 15:07:17,347 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2090, 2.5610, 3.5734, 2.1240, 1.7568, 3.7208, 0.6612, 2.2245], + device='cuda:1'), covar=tensor([0.1527, 0.1110, 0.0379, 0.2262, 0.3249, 0.0197, 0.3155, 0.1724], + device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0174, 0.0106, 0.0215, 0.0256, 0.0109, 0.0161, 0.0169], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 15:07:27,807 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 2.655e+02 3.191e+02 3.880e+02 8.243e+02, threshold=6.382e+02, percent-clipped=2.0 +2023-02-06 15:07:30,611 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 15:07:39,031 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5430, 1.9891, 2.9056, 2.2759, 2.7570, 2.3303, 2.0004, 1.4271], + device='cuda:1'), covar=tensor([0.4051, 0.4505, 0.1470, 0.3277, 0.2232, 0.2561, 0.1767, 0.4897], + device='cuda:1'), in_proj_covar=tensor([0.0895, 0.0896, 0.0742, 0.0869, 0.0949, 0.0827, 0.0710, 0.0778], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 15:07:41,535 INFO [train.py:901] (1/4) Epoch 14, batch 3700, loss[loss=0.2131, simple_loss=0.2992, pruned_loss=0.06353, over 8105.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3015, pruned_loss=0.07298, over 1612996.10 frames. ], batch size: 23, lr: 5.46e-03, grad_scale: 8.0 +2023-02-06 15:07:50,763 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:07:56,409 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:08:01,069 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108809.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:08:14,687 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6443, 1.4749, 1.5686, 1.2871, 0.9618, 1.3248, 1.6171, 1.3351], + device='cuda:1'), covar=tensor([0.0524, 0.1141, 0.1635, 0.1343, 0.0545, 0.1377, 0.0643, 0.0611], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0151, 0.0190, 0.0157, 0.0100, 0.0161, 0.0113, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 15:08:15,867 INFO [train.py:901] (1/4) Epoch 14, batch 3750, loss[loss=0.2175, simple_loss=0.2881, pruned_loss=0.07347, over 7797.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.301, pruned_loss=0.07296, over 1610893.71 frames. ], batch size: 19, lr: 5.46e-03, grad_scale: 8.0 +2023-02-06 15:08:37,495 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 2.414e+02 2.846e+02 3.664e+02 8.039e+02, threshold=5.692e+02, percent-clipped=5.0 +2023-02-06 15:08:51,949 INFO [train.py:901] (1/4) Epoch 14, batch 3800, loss[loss=0.2509, simple_loss=0.3283, pruned_loss=0.08669, over 8587.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3, pruned_loss=0.07222, over 1603249.46 frames. ], batch size: 31, lr: 5.46e-03, grad_scale: 8.0 +2023-02-06 15:09:12,257 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108909.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:09:21,323 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0648, 1.5955, 3.5454, 1.5000, 2.3774, 3.8409, 3.9421, 3.2952], + device='cuda:1'), covar=tensor([0.1041, 0.1675, 0.0285, 0.2142, 0.1059, 0.0277, 0.0479, 0.0607], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0303, 0.0266, 0.0296, 0.0283, 0.0246, 0.0367, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:09:26,888 INFO [train.py:901] (1/4) Epoch 14, batch 3850, loss[loss=0.2136, simple_loss=0.2984, pruned_loss=0.06439, over 8324.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.2995, pruned_loss=0.07213, over 1599041.49 frames. ], batch size: 25, lr: 5.46e-03, grad_scale: 16.0 +2023-02-06 15:09:33,935 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:09:35,490 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2835, 1.9332, 2.8605, 2.2708, 2.7294, 2.2186, 1.8846, 1.3531], + device='cuda:1'), covar=tensor([0.4635, 0.4454, 0.1470, 0.3051, 0.2076, 0.2492, 0.1755, 0.4875], + device='cuda:1'), in_proj_covar=tensor([0.0892, 0.0896, 0.0741, 0.0869, 0.0948, 0.0825, 0.0708, 0.0777], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 15:09:35,935 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 15:09:49,066 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.574e+02 3.020e+02 4.517e+02 9.725e+02, threshold=6.039e+02, percent-clipped=15.0 +2023-02-06 15:09:51,343 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7212, 2.0816, 1.6886, 2.5489, 1.1763, 1.3731, 1.7504, 2.2030], + device='cuda:1'), covar=tensor([0.0905, 0.0790, 0.1048, 0.0405, 0.1208, 0.1649, 0.1023, 0.0748], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0207, 0.0252, 0.0214, 0.0216, 0.0252, 0.0257, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 15:10:04,288 INFO [train.py:901] (1/4) Epoch 14, batch 3900, loss[loss=0.2031, simple_loss=0.2751, pruned_loss=0.06562, over 7932.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2986, pruned_loss=0.0716, over 1598282.64 frames. ], batch size: 20, lr: 5.46e-03, grad_scale: 16.0 +2023-02-06 15:10:07,765 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1695, 1.2147, 4.5234, 1.7160, 3.5677, 3.5268, 4.0578, 3.9786], + device='cuda:1'), covar=tensor([0.1020, 0.6822, 0.0833, 0.4502, 0.1987, 0.1670, 0.0961, 0.0901], + device='cuda:1'), in_proj_covar=tensor([0.0537, 0.0608, 0.0623, 0.0570, 0.0645, 0.0552, 0.0543, 0.0606], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 15:10:39,032 INFO [train.py:901] (1/4) Epoch 14, batch 3950, loss[loss=0.2039, simple_loss=0.2818, pruned_loss=0.06303, over 7184.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2995, pruned_loss=0.07201, over 1600284.04 frames. ], batch size: 16, lr: 5.46e-03, grad_scale: 16.0 +2023-02-06 15:10:56,309 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109054.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:10:56,952 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109055.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:10:59,103 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109058.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:11:00,258 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.734e+02 2.458e+02 2.966e+02 3.777e+02 8.079e+02, threshold=5.932e+02, percent-clipped=4.0 +2023-02-06 15:11:14,754 INFO [train.py:901] (1/4) Epoch 14, batch 4000, loss[loss=0.1896, simple_loss=0.2717, pruned_loss=0.05372, over 7655.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3008, pruned_loss=0.07255, over 1603842.35 frames. ], batch size: 19, lr: 5.46e-03, grad_scale: 16.0 +2023-02-06 15:11:17,682 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109083.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:11:50,550 INFO [train.py:901] (1/4) Epoch 14, batch 4050, loss[loss=0.2258, simple_loss=0.3134, pruned_loss=0.06912, over 8335.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3017, pruned_loss=0.0726, over 1608311.04 frames. ], batch size: 26, lr: 5.45e-03, grad_scale: 16.0 +2023-02-06 15:12:06,806 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:12:11,580 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-06 15:12:11,649 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.362e+02 2.684e+02 3.543e+02 7.215e+02, threshold=5.369e+02, percent-clipped=4.0 +2023-02-06 15:12:16,040 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:12:26,284 INFO [train.py:901] (1/4) Epoch 14, batch 4100, loss[loss=0.2234, simple_loss=0.3021, pruned_loss=0.07239, over 8553.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3022, pruned_loss=0.07244, over 1610530.06 frames. ], batch size: 31, lr: 5.45e-03, grad_scale: 16.0 +2023-02-06 15:12:33,980 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109190.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:12:50,498 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109212.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:13:02,496 INFO [train.py:901] (1/4) Epoch 14, batch 4150, loss[loss=0.2797, simple_loss=0.3346, pruned_loss=0.1124, over 6675.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3028, pruned_loss=0.07326, over 1610001.84 frames. ], batch size: 71, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:13:23,981 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.677e+02 3.078e+02 3.893e+02 8.547e+02, threshold=6.157e+02, percent-clipped=10.0 +2023-02-06 15:13:28,945 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109268.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:13:35,738 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 15:13:37,136 INFO [train.py:901] (1/4) Epoch 14, batch 4200, loss[loss=0.2379, simple_loss=0.3183, pruned_loss=0.07875, over 8575.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3023, pruned_loss=0.07288, over 1608800.91 frames. ], batch size: 31, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:13:44,010 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.98 vs. limit=5.0 +2023-02-06 15:13:59,678 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109310.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:14:00,998 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:14:01,570 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 15:14:14,547 INFO [train.py:901] (1/4) Epoch 14, batch 4250, loss[loss=0.2078, simple_loss=0.2981, pruned_loss=0.05876, over 8623.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3021, pruned_loss=0.07261, over 1607671.28 frames. ], batch size: 31, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:14:18,160 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109335.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:14:18,815 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:14:35,152 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.2583, 1.9012, 5.3923, 2.5116, 4.8282, 4.5514, 4.9796, 4.8502], + device='cuda:1'), covar=tensor([0.0512, 0.4498, 0.0456, 0.3163, 0.0970, 0.0803, 0.0454, 0.0471], + device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0599, 0.0617, 0.0559, 0.0633, 0.0544, 0.0535, 0.0596], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 15:14:35,685 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.488e+02 3.016e+02 3.845e+02 8.299e+02, threshold=6.033e+02, percent-clipped=4.0 +2023-02-06 15:14:48,673 INFO [train.py:901] (1/4) Epoch 14, batch 4300, loss[loss=0.2272, simple_loss=0.304, pruned_loss=0.07519, over 8232.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3022, pruned_loss=0.07276, over 1611010.28 frames. ], batch size: 22, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:14:56,019 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 15:14:58,772 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 15:15:01,894 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109399.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:15:24,494 INFO [train.py:901] (1/4) Epoch 14, batch 4350, loss[loss=0.2083, simple_loss=0.2828, pruned_loss=0.06689, over 7780.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3022, pruned_loss=0.0729, over 1610352.87 frames. ], batch size: 19, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:15:34,087 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 15:15:47,273 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 2.678e+02 3.270e+02 4.253e+02 1.326e+03, threshold=6.540e+02, percent-clipped=8.0 +2023-02-06 15:16:00,544 INFO [train.py:901] (1/4) Epoch 14, batch 4400, loss[loss=0.2819, simple_loss=0.347, pruned_loss=0.1084, over 8602.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3024, pruned_loss=0.07326, over 1615520.78 frames. ], batch size: 34, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:16:15,760 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 15:16:24,168 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109514.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:16:31,889 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:16:35,897 INFO [train.py:901] (1/4) Epoch 14, batch 4450, loss[loss=0.2861, simple_loss=0.3473, pruned_loss=0.1125, over 8455.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3009, pruned_loss=0.07284, over 1612758.78 frames. ], batch size: 27, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:16:49,285 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109549.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:16:55,365 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109556.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:16:58,638 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.456e+02 2.864e+02 3.608e+02 1.087e+03, threshold=5.728e+02, percent-clipped=4.0 +2023-02-06 15:17:11,132 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 15:17:12,488 INFO [train.py:901] (1/4) Epoch 14, batch 4500, loss[loss=0.191, simple_loss=0.2762, pruned_loss=0.05294, over 8101.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3011, pruned_loss=0.07295, over 1615534.96 frames. ], batch size: 21, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:17:24,282 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109597.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:17:47,001 INFO [train.py:901] (1/4) Epoch 14, batch 4550, loss[loss=0.2081, simple_loss=0.2757, pruned_loss=0.0702, over 7183.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3021, pruned_loss=0.07351, over 1614520.33 frames. ], batch size: 16, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:18:05,732 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109656.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:18:09,030 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.742e+02 2.640e+02 3.232e+02 4.162e+02 9.021e+02, threshold=6.464e+02, percent-clipped=8.0 +2023-02-06 15:18:16,730 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109671.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:18:21,369 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109677.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:18:23,261 INFO [train.py:901] (1/4) Epoch 14, batch 4600, loss[loss=0.2314, simple_loss=0.3053, pruned_loss=0.07872, over 8181.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2996, pruned_loss=0.07196, over 1611497.55 frames. ], batch size: 23, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:18:23,335 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109680.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:18:56,948 INFO [train.py:901] (1/4) Epoch 14, batch 4650, loss[loss=0.2516, simple_loss=0.324, pruned_loss=0.08961, over 8288.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3009, pruned_loss=0.07362, over 1605190.35 frames. ], batch size: 23, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:19:18,714 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.556e+02 3.032e+02 3.907e+02 9.020e+02, threshold=6.065e+02, percent-clipped=4.0 +2023-02-06 15:19:24,824 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109770.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:19:25,417 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109771.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:19:31,314 INFO [train.py:901] (1/4) Epoch 14, batch 4700, loss[loss=0.2613, simple_loss=0.3297, pruned_loss=0.09647, over 8318.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.2996, pruned_loss=0.07296, over 1602594.89 frames. ], batch size: 26, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:19:42,825 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109795.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:19:42,849 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109795.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:20:06,516 INFO [train.py:901] (1/4) Epoch 14, batch 4750, loss[loss=0.254, simple_loss=0.3271, pruned_loss=0.09049, over 8344.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3005, pruned_loss=0.07334, over 1604574.96 frames. ], batch size: 25, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:20:10,485 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 15:20:12,436 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 15:20:26,957 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.643e+02 3.166e+02 4.371e+02 1.104e+03, threshold=6.332e+02, percent-clipped=5.0 +2023-02-06 15:20:40,298 INFO [train.py:901] (1/4) Epoch 14, batch 4800, loss[loss=0.2704, simple_loss=0.3391, pruned_loss=0.1009, over 8364.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3009, pruned_loss=0.07358, over 1607947.83 frames. ], batch size: 24, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:20:56,506 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7336, 1.8151, 1.7173, 2.3252, 0.9703, 1.4895, 1.7869, 1.9455], + device='cuda:1'), covar=tensor([0.0728, 0.0865, 0.0916, 0.0378, 0.1196, 0.1400, 0.0756, 0.0759], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0207, 0.0253, 0.0212, 0.0214, 0.0251, 0.0256, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 15:21:03,935 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 15:21:14,269 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:21:16,045 INFO [train.py:901] (1/4) Epoch 14, batch 4850, loss[loss=0.181, simple_loss=0.2535, pruned_loss=0.05422, over 7425.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3, pruned_loss=0.07258, over 1605338.55 frames. ], batch size: 17, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:21:23,511 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109941.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:21:31,142 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109952.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:21:37,038 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.601e+02 2.400e+02 2.854e+02 3.344e+02 7.947e+02, threshold=5.708e+02, percent-clipped=2.0 +2023-02-06 15:21:38,878 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-06 15:21:46,631 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0181, 2.1608, 1.7852, 2.7891, 1.1554, 1.6628, 2.0120, 2.3301], + device='cuda:1'), covar=tensor([0.0757, 0.0893, 0.1033, 0.0368, 0.1243, 0.1402, 0.1021, 0.0800], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0208, 0.0254, 0.0212, 0.0214, 0.0252, 0.0256, 0.0215], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 15:21:49,882 INFO [train.py:901] (1/4) Epoch 14, batch 4900, loss[loss=0.1876, simple_loss=0.2582, pruned_loss=0.05856, over 7443.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3012, pruned_loss=0.07277, over 1608510.00 frames. ], batch size: 17, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:22:07,941 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9059, 1.8233, 2.9850, 1.5758, 2.3061, 3.2769, 3.2018, 2.8461], + device='cuda:1'), covar=tensor([0.0972, 0.1268, 0.0380, 0.1796, 0.0994, 0.0257, 0.0622, 0.0524], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0302, 0.0266, 0.0296, 0.0282, 0.0244, 0.0366, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 15:22:18,843 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0473, 1.5314, 3.4505, 1.5072, 2.3006, 3.8211, 3.7693, 3.2586], + device='cuda:1'), covar=tensor([0.0990, 0.1619, 0.0311, 0.2081, 0.1078, 0.0241, 0.0558, 0.0612], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0303, 0.0267, 0.0296, 0.0283, 0.0245, 0.0367, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 15:22:19,455 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110021.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:20,956 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3926, 2.7606, 3.2568, 1.6455, 3.4607, 1.9720, 1.5517, 2.1793], + device='cuda:1'), covar=tensor([0.0639, 0.0317, 0.0231, 0.0575, 0.0279, 0.0675, 0.0704, 0.0379], + device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0356, 0.0305, 0.0411, 0.0343, 0.0503, 0.0376, 0.0381], + device='cuda:1'), out_proj_covar=tensor([1.1645e-04, 9.6243e-05, 8.2705e-05, 1.1197e-04, 9.3788e-05, 1.4787e-04, + 1.0478e-04, 1.0453e-04], device='cuda:1') +2023-02-06 15:22:24,322 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110027.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:26,032 INFO [train.py:901] (1/4) Epoch 14, batch 4950, loss[loss=0.1927, simple_loss=0.2758, pruned_loss=0.05481, over 8298.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3014, pruned_loss=0.07295, over 1608672.32 frames. ], batch size: 23, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:22:30,965 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110035.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:41,741 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110051.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:42,418 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110052.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:45,147 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110056.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:48,328 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.585e+02 3.180e+02 4.032e+02 7.448e+02, threshold=6.360e+02, percent-clipped=3.0 +2023-02-06 15:22:49,148 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:58,267 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110076.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:23:00,768 INFO [train.py:901] (1/4) Epoch 14, batch 5000, loss[loss=0.2172, simple_loss=0.2982, pruned_loss=0.06803, over 8640.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3008, pruned_loss=0.0723, over 1610871.56 frames. ], batch size: 31, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:23:33,488 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:23:34,638 INFO [train.py:901] (1/4) Epoch 14, batch 5050, loss[loss=0.3106, simple_loss=0.3543, pruned_loss=0.1335, over 6660.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3013, pruned_loss=0.0732, over 1612036.39 frames. ], batch size: 71, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:23:36,756 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110133.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:23:38,745 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110136.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:23:43,180 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 15:23:57,272 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 2.889e+02 3.601e+02 4.263e+02 9.587e+02, threshold=7.203e+02, percent-clipped=6.0 +2023-02-06 15:24:09,950 INFO [train.py:901] (1/4) Epoch 14, batch 5100, loss[loss=0.2922, simple_loss=0.3551, pruned_loss=0.1147, over 7271.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3023, pruned_loss=0.07351, over 1617687.52 frames. ], batch size: 71, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:24:42,800 INFO [train.py:901] (1/4) Epoch 14, batch 5150, loss[loss=0.2047, simple_loss=0.2828, pruned_loss=0.06332, over 8281.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3025, pruned_loss=0.07378, over 1614343.84 frames. ], batch size: 23, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:25:05,059 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.455e+02 3.012e+02 3.817e+02 9.599e+02, threshold=6.024e+02, percent-clipped=2.0 +2023-02-06 15:25:17,476 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6635, 1.6980, 2.0960, 1.3408, 1.1872, 2.0918, 0.2815, 1.2534], + device='cuda:1'), covar=tensor([0.1771, 0.1271, 0.0419, 0.1800, 0.3521, 0.0442, 0.2444, 0.1837], + device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0173, 0.0105, 0.0214, 0.0258, 0.0111, 0.0160, 0.0170], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 15:25:19,324 INFO [train.py:901] (1/4) Epoch 14, batch 5200, loss[loss=0.2179, simple_loss=0.2934, pruned_loss=0.07116, over 8077.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.302, pruned_loss=0.07329, over 1613459.82 frames. ], batch size: 21, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:25:20,643 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1660, 4.1235, 3.7793, 1.7570, 3.7122, 3.7504, 3.7255, 3.3883], + device='cuda:1'), covar=tensor([0.0859, 0.0604, 0.1092, 0.5026, 0.0852, 0.1090, 0.1387, 0.0952], + device='cuda:1'), in_proj_covar=tensor([0.0480, 0.0397, 0.0401, 0.0495, 0.0393, 0.0398, 0.0387, 0.0346], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 15:25:25,792 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-06 15:25:27,240 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 +2023-02-06 15:25:33,781 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110301.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:25:39,477 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 15:25:41,058 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:25:47,126 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110321.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:25:53,178 INFO [train.py:901] (1/4) Epoch 14, batch 5250, loss[loss=0.2244, simple_loss=0.3063, pruned_loss=0.07125, over 8445.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3027, pruned_loss=0.07359, over 1617467.51 frames. ], batch size: 27, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:25:57,548 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:25:58,343 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110337.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:26:15,521 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.532e+02 3.204e+02 3.879e+02 8.466e+02, threshold=6.409e+02, percent-clipped=5.0 +2023-02-06 15:26:28,848 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:26:29,491 INFO [train.py:901] (1/4) Epoch 14, batch 5300, loss[loss=0.2689, simple_loss=0.3348, pruned_loss=0.1015, over 8346.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.303, pruned_loss=0.07425, over 1613440.52 frames. ], batch size: 26, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:26:37,622 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 15:26:39,498 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:26:48,919 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110406.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:26:56,332 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:27:03,329 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 +2023-02-06 15:27:05,004 INFO [train.py:901] (1/4) Epoch 14, batch 5350, loss[loss=0.237, simple_loss=0.3206, pruned_loss=0.07669, over 8506.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3028, pruned_loss=0.07389, over 1616764.95 frames. ], batch size: 28, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:27:25,506 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.452e+02 3.047e+02 3.791e+02 6.566e+02, threshold=6.094e+02, percent-clipped=2.0 +2023-02-06 15:27:32,802 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:27:36,875 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:27:38,761 INFO [train.py:901] (1/4) Epoch 14, batch 5400, loss[loss=0.2212, simple_loss=0.3036, pruned_loss=0.06939, over 8759.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3027, pruned_loss=0.07359, over 1619477.63 frames. ], batch size: 39, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:27:48,336 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110494.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:28:08,456 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110521.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:28:14,356 INFO [train.py:901] (1/4) Epoch 14, batch 5450, loss[loss=0.1969, simple_loss=0.2724, pruned_loss=0.06073, over 7920.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3019, pruned_loss=0.07307, over 1614540.42 frames. ], batch size: 20, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:28:30,474 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 15:28:34,934 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.429e+02 2.846e+02 3.589e+02 7.640e+02, threshold=5.692e+02, percent-clipped=1.0 +2023-02-06 15:28:39,770 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1573, 2.1254, 1.5811, 1.8501, 1.8187, 1.3618, 1.6551, 1.6052], + device='cuda:1'), covar=tensor([0.1430, 0.0424, 0.1299, 0.0564, 0.0724, 0.1573, 0.0946, 0.0892], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0235, 0.0322, 0.0297, 0.0298, 0.0328, 0.0344, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:28:47,544 INFO [train.py:901] (1/4) Epoch 14, batch 5500, loss[loss=0.2169, simple_loss=0.2999, pruned_loss=0.06695, over 8453.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.303, pruned_loss=0.07367, over 1617394.79 frames. ], batch size: 25, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:28:50,377 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4821, 2.8236, 1.9573, 2.2218, 2.3276, 1.7261, 2.1350, 2.1516], + device='cuda:1'), covar=tensor([0.1520, 0.0399, 0.1106, 0.0664, 0.0617, 0.1401, 0.0963, 0.0910], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0234, 0.0322, 0.0297, 0.0297, 0.0328, 0.0343, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:28:52,322 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:28:55,814 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:29:23,306 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110629.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:29:23,843 INFO [train.py:901] (1/4) Epoch 14, batch 5550, loss[loss=0.2107, simple_loss=0.2859, pruned_loss=0.06782, over 8241.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3018, pruned_loss=0.07333, over 1609730.94 frames. ], batch size: 22, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:29:28,292 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.94 vs. limit=5.0 +2023-02-06 15:29:34,006 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:29:35,490 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0966, 4.1948, 2.6163, 2.7562, 3.0842, 2.2581, 2.8959, 3.0442], + device='cuda:1'), covar=tensor([0.1546, 0.0239, 0.0841, 0.0686, 0.0696, 0.1241, 0.0987, 0.0996], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0235, 0.0323, 0.0298, 0.0298, 0.0329, 0.0345, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:29:44,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.692e+02 2.421e+02 3.120e+02 3.692e+02 1.093e+03, threshold=6.240e+02, percent-clipped=9.0 +2023-02-06 15:29:47,077 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110665.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:29:56,916 INFO [train.py:901] (1/4) Epoch 14, batch 5600, loss[loss=0.2221, simple_loss=0.2993, pruned_loss=0.07242, over 8485.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3027, pruned_loss=0.07412, over 1613780.49 frames. ], batch size: 28, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:29:56,986 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110680.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:30:30,943 INFO [train.py:901] (1/4) Epoch 14, batch 5650, loss[loss=0.2635, simple_loss=0.3343, pruned_loss=0.09637, over 7180.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3022, pruned_loss=0.07345, over 1615686.52 frames. ], batch size: 71, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:30:33,768 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 15:30:46,525 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9618, 1.5786, 1.2024, 1.4286, 1.3109, 1.0615, 1.1631, 1.2015], + device='cuda:1'), covar=tensor([0.1050, 0.0512, 0.1265, 0.0612, 0.0792, 0.1585, 0.0992, 0.0775], + device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0233, 0.0319, 0.0294, 0.0294, 0.0324, 0.0340, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:30:47,219 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:30:49,877 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:30:54,024 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:30:54,499 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.502e+02 3.092e+02 3.638e+02 5.778e+02, threshold=6.185e+02, percent-clipped=0.0 +2023-02-06 15:30:55,475 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4622, 1.7064, 2.7491, 1.3052, 1.9589, 1.7882, 1.5203, 1.9648], + device='cuda:1'), covar=tensor([0.1904, 0.2476, 0.0800, 0.4288, 0.1737, 0.3187, 0.2115, 0.2174], + device='cuda:1'), in_proj_covar=tensor([0.0499, 0.0552, 0.0540, 0.0598, 0.0621, 0.0565, 0.0490, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 15:30:56,821 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2995, 1.4400, 1.3403, 1.8213, 0.7487, 1.1007, 1.3448, 1.4941], + device='cuda:1'), covar=tensor([0.0891, 0.0821, 0.1101, 0.0540, 0.1140, 0.1534, 0.0789, 0.0777], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0204, 0.0250, 0.0211, 0.0211, 0.0248, 0.0252, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 15:31:04,326 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110775.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:04,551 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.66 vs. limit=5.0 +2023-02-06 15:31:05,711 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110777.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:07,496 INFO [train.py:901] (1/4) Epoch 14, batch 5700, loss[loss=0.2335, simple_loss=0.3163, pruned_loss=0.07535, over 8359.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3041, pruned_loss=0.07403, over 1622803.11 frames. ], batch size: 24, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:31:07,673 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110780.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:12,399 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2858, 1.6937, 3.3591, 1.6012, 2.4603, 3.7326, 3.6857, 3.2286], + device='cuda:1'), covar=tensor([0.0922, 0.1633, 0.0369, 0.2063, 0.1111, 0.0231, 0.0563, 0.0528], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0303, 0.0268, 0.0295, 0.0282, 0.0243, 0.0368, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 15:31:12,424 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2192, 1.6703, 1.7115, 1.5494, 1.0345, 1.6031, 1.8004, 1.9027], + device='cuda:1'), covar=tensor([0.0487, 0.1171, 0.1656, 0.1330, 0.0652, 0.1468, 0.0697, 0.0532], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0157, 0.0102, 0.0162, 0.0114, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 15:31:17,893 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110795.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:22,879 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:40,944 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 15:31:41,628 INFO [train.py:901] (1/4) Epoch 14, batch 5750, loss[loss=0.1802, simple_loss=0.2774, pruned_loss=0.04153, over 8189.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3021, pruned_loss=0.07273, over 1619090.32 frames. ], batch size: 23, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:31:51,512 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110843.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:54,986 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:32:04,403 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.447e+02 3.019e+02 3.853e+02 7.521e+02, threshold=6.038e+02, percent-clipped=3.0 +2023-02-06 15:32:10,139 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:32:14,045 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110873.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:32:18,669 INFO [train.py:901] (1/4) Epoch 14, batch 5800, loss[loss=0.2241, simple_loss=0.3025, pruned_loss=0.07284, over 8244.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.301, pruned_loss=0.07211, over 1619567.25 frames. ], batch size: 22, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:32:22,958 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:32:53,262 INFO [train.py:901] (1/4) Epoch 14, batch 5850, loss[loss=0.1887, simple_loss=0.2607, pruned_loss=0.05834, over 7536.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3004, pruned_loss=0.07166, over 1621600.13 frames. ], batch size: 18, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:33:14,140 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.687e+02 2.498e+02 3.098e+02 4.112e+02 1.106e+03, threshold=6.195e+02, percent-clipped=10.0 +2023-02-06 15:33:15,331 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 15:33:22,880 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110973.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:33:25,739 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110976.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:33:28,977 INFO [train.py:901] (1/4) Epoch 14, batch 5900, loss[loss=0.274, simple_loss=0.3397, pruned_loss=0.1041, over 8087.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2997, pruned_loss=0.07179, over 1616238.16 frames. ], batch size: 21, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:33:40,957 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 15:33:51,006 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7881, 1.7586, 2.4745, 1.6922, 1.2286, 2.5912, 0.4896, 1.3198], + device='cuda:1'), covar=tensor([0.2194, 0.1427, 0.0409, 0.1635, 0.3465, 0.0263, 0.2638, 0.1951], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0175, 0.0107, 0.0216, 0.0260, 0.0112, 0.0161, 0.0172], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 15:33:54,429 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111016.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:03,738 INFO [train.py:901] (1/4) Epoch 14, batch 5950, loss[loss=0.2696, simple_loss=0.3292, pruned_loss=0.105, over 6822.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3011, pruned_loss=0.07225, over 1613786.98 frames. ], batch size: 71, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:34:07,788 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111036.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:08,695 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.18 vs. limit=5.0 +2023-02-06 15:34:11,072 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111041.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:17,718 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111051.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:20,640 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 15:34:24,231 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.648e+02 3.047e+02 4.016e+02 7.772e+02, threshold=6.093e+02, percent-clipped=5.0 +2023-02-06 15:34:24,456 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111061.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:34,726 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111076.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:38,002 INFO [train.py:901] (1/4) Epoch 14, batch 6000, loss[loss=0.1972, simple_loss=0.27, pruned_loss=0.06225, over 7928.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3007, pruned_loss=0.07226, over 1614362.34 frames. ], batch size: 20, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:34:38,002 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 15:34:50,553 INFO [train.py:935] (1/4) Epoch 14, validation: loss=0.1818, simple_loss=0.2816, pruned_loss=0.04094, over 944034.00 frames. +2023-02-06 15:34:50,554 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 15:34:56,289 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111088.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:35:03,697 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111098.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:35:27,233 INFO [train.py:901] (1/4) Epoch 14, batch 6050, loss[loss=0.2326, simple_loss=0.3103, pruned_loss=0.07745, over 8075.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3012, pruned_loss=0.07211, over 1615685.25 frames. ], batch size: 21, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:35:45,547 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6232, 1.4815, 2.8675, 1.3368, 2.1221, 3.1028, 3.1678, 2.6197], + device='cuda:1'), covar=tensor([0.1120, 0.1513, 0.0366, 0.2038, 0.0857, 0.0284, 0.0557, 0.0666], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0304, 0.0266, 0.0294, 0.0282, 0.0243, 0.0369, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 15:35:49,431 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.663e+02 2.432e+02 2.876e+02 3.526e+02 5.542e+02, threshold=5.752e+02, percent-clipped=0.0 +2023-02-06 15:36:01,587 INFO [train.py:901] (1/4) Epoch 14, batch 6100, loss[loss=0.2175, simple_loss=0.282, pruned_loss=0.07644, over 7793.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3012, pruned_loss=0.07282, over 1614885.39 frames. ], batch size: 19, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:36:15,956 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 15:36:24,938 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:36:37,111 INFO [train.py:901] (1/4) Epoch 14, batch 6150, loss[loss=0.1932, simple_loss=0.2771, pruned_loss=0.05466, over 7933.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3014, pruned_loss=0.07307, over 1617227.60 frames. ], batch size: 20, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:36:37,203 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111230.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:36:46,726 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111243.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:36:59,331 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.474e+02 3.213e+02 4.029e+02 8.079e+02, threshold=6.426e+02, percent-clipped=5.0 +2023-02-06 15:37:11,858 INFO [train.py:901] (1/4) Epoch 14, batch 6200, loss[loss=0.205, simple_loss=0.2934, pruned_loss=0.05826, over 8561.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3021, pruned_loss=0.07369, over 1618140.44 frames. ], batch size: 39, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:37:13,480 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=5.19 vs. limit=5.0 +2023-02-06 15:37:38,513 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111320.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:37:45,101 INFO [train.py:901] (1/4) Epoch 14, batch 6250, loss[loss=0.1687, simple_loss=0.2562, pruned_loss=0.04061, over 7579.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.301, pruned_loss=0.07304, over 1612742.43 frames. ], batch size: 18, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:37:54,003 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 15:37:55,172 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111344.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:37:55,812 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111345.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:38:08,436 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.286e+02 2.818e+02 3.691e+02 1.208e+03, threshold=5.637e+02, percent-clipped=2.0 +2023-02-06 15:38:13,378 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111369.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:38:20,506 INFO [train.py:901] (1/4) Epoch 14, batch 6300, loss[loss=0.2016, simple_loss=0.286, pruned_loss=0.05858, over 8090.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3014, pruned_loss=0.0736, over 1611834.85 frames. ], batch size: 21, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:38:55,235 INFO [train.py:901] (1/4) Epoch 14, batch 6350, loss[loss=0.1986, simple_loss=0.2789, pruned_loss=0.05916, over 7663.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3013, pruned_loss=0.07365, over 1611740.60 frames. ], batch size: 19, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:38:56,130 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0888, 3.7109, 2.1826, 2.7871, 2.7555, 1.9809, 2.6969, 3.0304], + device='cuda:1'), covar=tensor([0.1577, 0.0321, 0.1093, 0.0806, 0.0723, 0.1293, 0.1039, 0.1110], + device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0232, 0.0323, 0.0293, 0.0298, 0.0326, 0.0341, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:38:58,925 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:38:58,978 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5398, 2.0717, 3.4275, 1.2496, 2.6827, 2.0278, 1.5125, 2.4271], + device='cuda:1'), covar=tensor([0.1783, 0.2535, 0.0808, 0.4380, 0.1619, 0.3075, 0.2182, 0.2467], + device='cuda:1'), in_proj_covar=tensor([0.0499, 0.0553, 0.0539, 0.0601, 0.0625, 0.0566, 0.0495, 0.0621], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 15:39:17,455 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 2.340e+02 2.891e+02 3.552e+02 9.934e+02, threshold=5.783e+02, percent-clipped=8.0 +2023-02-06 15:39:23,086 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111469.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:39:30,957 INFO [train.py:901] (1/4) Epoch 14, batch 6400, loss[loss=0.2035, simple_loss=0.2752, pruned_loss=0.06594, over 7517.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3004, pruned_loss=0.07305, over 1613458.58 frames. ], batch size: 18, lr: 5.40e-03, grad_scale: 8.0 +2023-02-06 15:39:32,195 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 15:39:35,213 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111486.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:39:40,674 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111494.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:40:05,256 INFO [train.py:901] (1/4) Epoch 14, batch 6450, loss[loss=0.2151, simple_loss=0.2891, pruned_loss=0.07049, over 8195.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2996, pruned_loss=0.07202, over 1613062.21 frames. ], batch size: 23, lr: 5.40e-03, grad_scale: 8.0 +2023-02-06 15:40:26,315 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.418e+02 3.184e+02 3.807e+02 1.482e+03, threshold=6.367e+02, percent-clipped=8.0 +2023-02-06 15:40:39,068 INFO [train.py:901] (1/4) Epoch 14, batch 6500, loss[loss=0.1681, simple_loss=0.2437, pruned_loss=0.04625, over 7688.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2999, pruned_loss=0.07186, over 1613899.63 frames. ], batch size: 18, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:40:44,376 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:40:55,099 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111601.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:41:12,089 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111626.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:41:14,602 INFO [train.py:901] (1/4) Epoch 14, batch 6550, loss[loss=0.2445, simple_loss=0.3286, pruned_loss=0.08024, over 8372.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3015, pruned_loss=0.07279, over 1614155.12 frames. ], batch size: 24, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:41:24,412 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 15:41:35,853 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 2.515e+02 3.055e+02 3.900e+02 7.605e+02, threshold=6.110e+02, percent-clipped=3.0 +2023-02-06 15:41:43,337 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 15:41:47,960 INFO [train.py:901] (1/4) Epoch 14, batch 6600, loss[loss=0.1904, simple_loss=0.2786, pruned_loss=0.05108, over 8255.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3032, pruned_loss=0.07383, over 1615420.33 frames. ], batch size: 24, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:41:55,628 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:42:03,622 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111702.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:42:14,397 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111716.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:42:23,465 INFO [train.py:901] (1/4) Epoch 14, batch 6650, loss[loss=0.2151, simple_loss=0.2926, pruned_loss=0.06878, over 8237.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3024, pruned_loss=0.07338, over 1614866.62 frames. ], batch size: 22, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:42:45,229 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.391e+02 3.105e+02 3.860e+02 7.189e+02, threshold=6.209e+02, percent-clipped=3.0 +2023-02-06 15:42:48,091 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9162, 2.2339, 1.8141, 2.8003, 1.3688, 1.4515, 1.8983, 2.3579], + device='cuda:1'), covar=tensor([0.0815, 0.0932, 0.1022, 0.0389, 0.1253, 0.1636, 0.1068, 0.0880], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0207, 0.0256, 0.0215, 0.0216, 0.0253, 0.0261, 0.0217], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 15:42:49,653 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 15:42:51,483 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0336, 2.4893, 2.8764, 1.4277, 3.0557, 1.7183, 1.4557, 1.8851], + device='cuda:1'), covar=tensor([0.0755, 0.0328, 0.0255, 0.0655, 0.0362, 0.0706, 0.0789, 0.0485], + device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0354, 0.0302, 0.0405, 0.0339, 0.0493, 0.0369, 0.0376], + device='cuda:1'), out_proj_covar=tensor([1.1356e-04, 9.5802e-05, 8.1746e-05, 1.1006e-04, 9.2653e-05, 1.4440e-04, + 1.0266e-04, 1.0282e-04], device='cuda:1') +2023-02-06 15:42:57,304 INFO [train.py:901] (1/4) Epoch 14, batch 6700, loss[loss=0.2359, simple_loss=0.3269, pruned_loss=0.0725, over 8315.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3024, pruned_loss=0.07378, over 1614176.35 frames. ], batch size: 25, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:43:16,620 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111809.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:43:32,519 INFO [train.py:901] (1/4) Epoch 14, batch 6750, loss[loss=0.2714, simple_loss=0.3478, pruned_loss=0.09749, over 8573.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3013, pruned_loss=0.07311, over 1612045.63 frames. ], batch size: 49, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:43:32,588 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111830.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:43:54,030 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.559e+02 3.020e+02 4.182e+02 1.269e+03, threshold=6.039e+02, percent-clipped=6.0 +2023-02-06 15:44:02,412 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 15:44:02,884 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 15:44:07,192 INFO [train.py:901] (1/4) Epoch 14, batch 6800, loss[loss=0.2423, simple_loss=0.3228, pruned_loss=0.08084, over 8288.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3016, pruned_loss=0.07328, over 1608309.07 frames. ], batch size: 23, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:44:40,418 INFO [train.py:901] (1/4) Epoch 14, batch 6850, loss[loss=0.2379, simple_loss=0.2932, pruned_loss=0.09125, over 7723.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3016, pruned_loss=0.07341, over 1609538.65 frames. ], batch size: 18, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:44:51,192 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 15:44:52,060 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111945.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:45:01,413 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111958.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:45:03,875 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.542e+02 3.126e+02 4.226e+02 8.027e+02, threshold=6.251e+02, percent-clipped=7.0 +2023-02-06 15:45:16,602 INFO [train.py:901] (1/4) Epoch 14, batch 6900, loss[loss=0.2702, simple_loss=0.3239, pruned_loss=0.1083, over 8254.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.302, pruned_loss=0.07337, over 1609867.65 frames. ], batch size: 22, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:45:18,708 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111983.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:45:50,586 INFO [train.py:901] (1/4) Epoch 14, batch 6950, loss[loss=0.2159, simple_loss=0.2927, pruned_loss=0.06951, over 7801.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3028, pruned_loss=0.07386, over 1612910.89 frames. ], batch size: 20, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:45:58,672 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 15:46:13,926 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.444e+02 2.419e+02 2.987e+02 3.531e+02 6.552e+02, threshold=5.974e+02, percent-clipped=1.0 +2023-02-06 15:46:25,966 INFO [train.py:901] (1/4) Epoch 14, batch 7000, loss[loss=0.2104, simple_loss=0.2798, pruned_loss=0.07046, over 6897.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3022, pruned_loss=0.07373, over 1604280.54 frames. ], batch size: 15, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:46:28,779 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112084.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:46:28,799 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9539, 1.5890, 3.3172, 1.3962, 2.3390, 3.6124, 3.6892, 3.0862], + device='cuda:1'), covar=tensor([0.1036, 0.1541, 0.0324, 0.2058, 0.0935, 0.0226, 0.0429, 0.0604], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0306, 0.0268, 0.0298, 0.0284, 0.0244, 0.0369, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:46:41,425 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8101, 3.8134, 3.4496, 1.8089, 3.3826, 3.4063, 3.4331, 3.1963], + device='cuda:1'), covar=tensor([0.1007, 0.0786, 0.1207, 0.4880, 0.1103, 0.1154, 0.1530, 0.0883], + device='cuda:1'), in_proj_covar=tensor([0.0485, 0.0397, 0.0403, 0.0500, 0.0396, 0.0400, 0.0390, 0.0346], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 15:46:59,921 INFO [train.py:901] (1/4) Epoch 14, batch 7050, loss[loss=0.2196, simple_loss=0.2952, pruned_loss=0.07197, over 8445.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.302, pruned_loss=0.07356, over 1608476.19 frames. ], batch size: 27, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:47:15,926 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112153.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:47:21,802 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.608e+02 3.153e+02 4.211e+02 1.237e+03, threshold=6.307e+02, percent-clipped=12.0 +2023-02-06 15:47:35,266 INFO [train.py:901] (1/4) Epoch 14, batch 7100, loss[loss=0.2515, simple_loss=0.3204, pruned_loss=0.09135, over 8527.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3025, pruned_loss=0.07356, over 1613547.11 frames. ], batch size: 39, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:47:50,288 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112201.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:47:59,152 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 15:48:07,018 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5715, 2.6663, 1.9294, 2.3601, 2.1976, 1.4543, 2.0765, 2.3070], + device='cuda:1'), covar=tensor([0.1496, 0.0426, 0.1000, 0.0588, 0.0735, 0.1528, 0.1012, 0.0916], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0234, 0.0323, 0.0298, 0.0301, 0.0329, 0.0345, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:48:07,693 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112226.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:48:10,128 INFO [train.py:901] (1/4) Epoch 14, batch 7150, loss[loss=0.228, simple_loss=0.314, pruned_loss=0.07103, over 8188.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3019, pruned_loss=0.07292, over 1612850.55 frames. ], batch size: 23, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:48:16,272 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3412, 1.6236, 4.5418, 1.7808, 3.9738, 3.7229, 4.0737, 3.9631], + device='cuda:1'), covar=tensor([0.0561, 0.4137, 0.0525, 0.3617, 0.1108, 0.0978, 0.0560, 0.0608], + device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0599, 0.0625, 0.0566, 0.0638, 0.0550, 0.0542, 0.0603], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 15:48:31,544 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.371e+02 2.859e+02 3.664e+02 7.587e+02, threshold=5.717e+02, percent-clipped=3.0 +2023-02-06 15:48:35,315 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-06 15:48:35,572 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112268.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:48:43,212 INFO [train.py:901] (1/4) Epoch 14, batch 7200, loss[loss=0.221, simple_loss=0.3023, pruned_loss=0.06979, over 8559.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.301, pruned_loss=0.07242, over 1614574.66 frames. ], batch size: 31, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:49:18,426 INFO [train.py:901] (1/4) Epoch 14, batch 7250, loss[loss=0.1927, simple_loss=0.261, pruned_loss=0.06222, over 7546.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3013, pruned_loss=0.07258, over 1617509.57 frames. ], batch size: 18, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:49:39,912 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.708e+02 3.212e+02 3.989e+02 8.387e+02, threshold=6.424e+02, percent-clipped=5.0 +2023-02-06 15:49:52,045 INFO [train.py:901] (1/4) Epoch 14, batch 7300, loss[loss=0.2028, simple_loss=0.2922, pruned_loss=0.05667, over 8460.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3021, pruned_loss=0.07337, over 1620852.39 frames. ], batch size: 25, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:49:54,806 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8267, 1.8413, 2.4489, 1.5775, 1.2155, 2.4534, 0.3290, 1.2474], + device='cuda:1'), covar=tensor([0.2319, 0.1541, 0.0353, 0.1954, 0.3962, 0.0359, 0.3135, 0.1967], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0175, 0.0107, 0.0215, 0.0256, 0.0111, 0.0161, 0.0173], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 15:50:11,924 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112407.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:50:15,499 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 15:50:24,101 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5007, 2.7760, 1.8122, 2.2071, 2.2919, 1.4731, 2.0218, 2.1858], + device='cuda:1'), covar=tensor([0.1502, 0.0340, 0.1148, 0.0653, 0.0719, 0.1541, 0.1032, 0.0900], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0235, 0.0326, 0.0299, 0.0301, 0.0329, 0.0346, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:50:25,903 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-06 15:50:26,701 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112428.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:50:28,013 INFO [train.py:901] (1/4) Epoch 14, batch 7350, loss[loss=0.2378, simple_loss=0.3115, pruned_loss=0.08202, over 8463.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3022, pruned_loss=0.07372, over 1616688.05 frames. ], batch size: 27, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:50:35,536 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2112, 1.5032, 4.6307, 2.2456, 2.5690, 5.1649, 5.2920, 4.4820], + device='cuda:1'), covar=tensor([0.1094, 0.1707, 0.0226, 0.1706, 0.0964, 0.0160, 0.0367, 0.0556], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0309, 0.0270, 0.0300, 0.0288, 0.0247, 0.0373, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 15:50:40,030 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 15:50:40,253 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5176, 1.9627, 2.9346, 1.3066, 2.2294, 1.6917, 1.7153, 1.9860], + device='cuda:1'), covar=tensor([0.1830, 0.2241, 0.0846, 0.4164, 0.1644, 0.3161, 0.1945, 0.2316], + device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0552, 0.0535, 0.0597, 0.0621, 0.0565, 0.0491, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 15:50:49,907 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 2.462e+02 2.972e+02 3.682e+02 1.093e+03, threshold=5.943e+02, percent-clipped=5.0 +2023-02-06 15:50:59,946 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 15:51:02,056 INFO [train.py:901] (1/4) Epoch 14, batch 7400, loss[loss=0.2041, simple_loss=0.2963, pruned_loss=0.05589, over 8333.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3019, pruned_loss=0.07322, over 1617069.23 frames. ], batch size: 25, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:51:10,094 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-02-06 15:51:32,629 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112524.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:51:37,189 INFO [train.py:901] (1/4) Epoch 14, batch 7450, loss[loss=0.1796, simple_loss=0.2698, pruned_loss=0.04467, over 8137.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3025, pruned_loss=0.0738, over 1609674.94 frames. ], batch size: 22, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:51:41,785 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 15:51:46,958 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:51:51,013 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112549.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:51:59,933 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.473e+02 3.112e+02 3.710e+02 6.215e+02, threshold=6.224e+02, percent-clipped=1.0 +2023-02-06 15:52:13,223 INFO [train.py:901] (1/4) Epoch 14, batch 7500, loss[loss=0.225, simple_loss=0.3189, pruned_loss=0.0655, over 8637.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3016, pruned_loss=0.07307, over 1604767.60 frames. ], batch size: 34, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:52:13,407 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112580.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:52:23,802 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112595.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:52:47,712 INFO [train.py:901] (1/4) Epoch 14, batch 7550, loss[loss=0.2224, simple_loss=0.3019, pruned_loss=0.07145, over 8759.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3008, pruned_loss=0.07278, over 1606762.31 frames. ], batch size: 30, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:52:51,319 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:53:11,225 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.402e+02 2.890e+02 3.643e+02 7.164e+02, threshold=5.781e+02, percent-clipped=3.0 +2023-02-06 15:53:23,631 INFO [train.py:901] (1/4) Epoch 14, batch 7600, loss[loss=0.2237, simple_loss=0.307, pruned_loss=0.07015, over 8024.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3002, pruned_loss=0.0718, over 1610928.73 frames. ], batch size: 22, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:53:40,424 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1414, 1.7937, 2.0203, 1.8518, 1.2325, 1.8327, 2.4931, 2.4800], + device='cuda:1'), covar=tensor([0.0383, 0.1167, 0.1581, 0.1266, 0.0559, 0.1336, 0.0565, 0.0521], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0152, 0.0190, 0.0155, 0.0101, 0.0161, 0.0115, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 15:53:57,110 INFO [train.py:901] (1/4) Epoch 14, batch 7650, loss[loss=0.2072, simple_loss=0.2816, pruned_loss=0.06635, over 7251.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3006, pruned_loss=0.07188, over 1609813.06 frames. ], batch size: 16, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:54:11,086 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112751.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:54:18,467 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.546e+02 2.941e+02 3.649e+02 7.123e+02, threshold=5.882e+02, percent-clipped=5.0 +2023-02-06 15:54:32,364 INFO [train.py:901] (1/4) Epoch 14, batch 7700, loss[loss=0.1806, simple_loss=0.2747, pruned_loss=0.04327, over 7925.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3, pruned_loss=0.07203, over 1605032.79 frames. ], batch size: 20, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:54:44,372 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 15:54:45,559 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112799.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:54:52,888 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 15:55:03,122 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:55:06,903 INFO [train.py:901] (1/4) Epoch 14, batch 7750, loss[loss=0.2297, simple_loss=0.3074, pruned_loss=0.07603, over 8249.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3005, pruned_loss=0.07184, over 1605756.69 frames. ], batch size: 24, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:55:28,266 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.453e+02 3.172e+02 4.245e+02 8.131e+02, threshold=6.343e+02, percent-clipped=10.0 +2023-02-06 15:55:30,962 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112866.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:55:41,030 INFO [train.py:901] (1/4) Epoch 14, batch 7800, loss[loss=0.2581, simple_loss=0.3361, pruned_loss=0.09001, over 8551.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3002, pruned_loss=0.07136, over 1609521.15 frames. ], batch size: 39, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:56:12,169 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112924.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:56:16,243 INFO [train.py:901] (1/4) Epoch 14, batch 7850, loss[loss=0.2055, simple_loss=0.2861, pruned_loss=0.06246, over 8028.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3012, pruned_loss=0.07219, over 1612038.90 frames. ], batch size: 22, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:56:22,361 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:56:22,439 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:56:29,119 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5588, 1.9305, 2.1039, 1.2645, 2.1096, 1.4255, 0.5309, 1.8439], + device='cuda:1'), covar=tensor([0.0442, 0.0260, 0.0212, 0.0411, 0.0317, 0.0710, 0.0669, 0.0219], + device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0355, 0.0302, 0.0407, 0.0342, 0.0495, 0.0368, 0.0380], + device='cuda:1'), out_proj_covar=tensor([1.1508e-04, 9.5920e-05, 8.1591e-05, 1.1046e-04, 9.3161e-05, 1.4483e-04, + 1.0230e-04, 1.0379e-04], device='cuda:1') +2023-02-06 15:56:37,784 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 2.504e+02 3.067e+02 3.726e+02 7.698e+02, threshold=6.135e+02, percent-clipped=2.0 +2023-02-06 15:56:49,065 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:56:49,649 INFO [train.py:901] (1/4) Epoch 14, batch 7900, loss[loss=0.232, simple_loss=0.316, pruned_loss=0.07396, over 8322.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3017, pruned_loss=0.07217, over 1615074.87 frames. ], batch size: 25, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:57:22,239 INFO [train.py:901] (1/4) Epoch 14, batch 7950, loss[loss=0.2264, simple_loss=0.3166, pruned_loss=0.06806, over 8351.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3023, pruned_loss=0.07241, over 1621247.29 frames. ], batch size: 24, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:57:28,310 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113039.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:57:38,045 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113054.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:57:43,246 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.654e+02 3.191e+02 4.041e+02 1.304e+03, threshold=6.382e+02, percent-clipped=5.0 +2023-02-06 15:57:55,254 INFO [train.py:901] (1/4) Epoch 14, batch 8000, loss[loss=0.227, simple_loss=0.3028, pruned_loss=0.07561, over 7930.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3029, pruned_loss=0.0732, over 1623149.19 frames. ], batch size: 20, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:58:04,698 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:58:23,452 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113122.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:58:28,564 INFO [train.py:901] (1/4) Epoch 14, batch 8050, loss[loss=0.2104, simple_loss=0.2773, pruned_loss=0.07174, over 7263.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3013, pruned_loss=0.07343, over 1594109.26 frames. ], batch size: 16, lr: 5.36e-03, grad_scale: 16.0 +2023-02-06 15:58:38,807 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7565, 2.2651, 4.5144, 1.4450, 3.3167, 2.3262, 1.9481, 3.1380], + device='cuda:1'), covar=tensor([0.1774, 0.2389, 0.0751, 0.4377, 0.1523, 0.2914, 0.2057, 0.2134], + device='cuda:1'), in_proj_covar=tensor([0.0500, 0.0554, 0.0538, 0.0606, 0.0626, 0.0569, 0.0495, 0.0625], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 15:58:40,125 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113147.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:58:49,778 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.323e+02 2.856e+02 3.288e+02 8.076e+02, threshold=5.712e+02, percent-clipped=1.0 +2023-02-06 15:59:01,666 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 15:59:06,246 INFO [train.py:901] (1/4) Epoch 15, batch 0, loss[loss=0.242, simple_loss=0.3087, pruned_loss=0.08762, over 8241.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3087, pruned_loss=0.08762, over 8241.00 frames. ], batch size: 22, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 15:59:06,246 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 15:59:17,269 INFO [train.py:935] (1/4) Epoch 15, validation: loss=0.1825, simple_loss=0.283, pruned_loss=0.04098, over 944034.00 frames. +2023-02-06 15:59:17,270 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 15:59:32,304 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 15:59:48,307 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 +2023-02-06 15:59:49,085 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 15:59:51,456 INFO [train.py:901] (1/4) Epoch 15, batch 50, loss[loss=0.225, simple_loss=0.3089, pruned_loss=0.07053, over 8244.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3019, pruned_loss=0.06903, over 370578.86 frames. ], batch size: 22, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:00:08,699 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 16:00:21,171 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113252.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:00:27,789 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.520e+02 2.549e+02 3.077e+02 3.582e+02 9.445e+02, threshold=6.153e+02, percent-clipped=5.0 +2023-02-06 16:00:28,497 INFO [train.py:901] (1/4) Epoch 15, batch 100, loss[loss=0.2043, simple_loss=0.273, pruned_loss=0.0678, over 7556.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3031, pruned_loss=0.0729, over 647668.76 frames. ], batch size: 18, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:00:29,918 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 16:00:39,303 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6265, 5.6566, 5.0115, 2.1460, 5.0179, 5.4562, 5.2898, 5.2610], + device='cuda:1'), covar=tensor([0.0583, 0.0416, 0.1034, 0.5160, 0.0827, 0.0785, 0.1057, 0.0553], + device='cuda:1'), in_proj_covar=tensor([0.0484, 0.0404, 0.0403, 0.0503, 0.0402, 0.0400, 0.0393, 0.0352], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:00:41,966 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:00:50,249 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:00,334 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113310.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:02,086 INFO [train.py:901] (1/4) Epoch 15, batch 150, loss[loss=0.2302, simple_loss=0.3016, pruned_loss=0.07936, over 8084.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3022, pruned_loss=0.07236, over 865066.56 frames. ], batch size: 21, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:01:06,868 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113320.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:17,348 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113335.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:28,778 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113350.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:32,820 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4890, 1.6681, 1.8570, 1.0889, 1.9758, 1.3095, 0.5509, 1.6965], + device='cuda:1'), covar=tensor([0.0435, 0.0266, 0.0212, 0.0404, 0.0261, 0.0653, 0.0627, 0.0194], + device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0358, 0.0304, 0.0408, 0.0343, 0.0497, 0.0370, 0.0380], + device='cuda:1'), out_proj_covar=tensor([1.1544e-04, 9.6722e-05, 8.2157e-05, 1.1080e-04, 9.3424e-05, 1.4533e-04, + 1.0253e-04, 1.0406e-04], device='cuda:1') +2023-02-06 16:01:37,328 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 2.511e+02 3.032e+02 4.146e+02 1.005e+03, threshold=6.064e+02, percent-clipped=3.0 +2023-02-06 16:01:38,024 INFO [train.py:901] (1/4) Epoch 15, batch 200, loss[loss=0.2311, simple_loss=0.3106, pruned_loss=0.07579, over 8351.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.303, pruned_loss=0.07292, over 1033025.01 frames. ], batch size: 26, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:01:46,249 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113375.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:02:01,348 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:02:11,079 INFO [train.py:901] (1/4) Epoch 15, batch 250, loss[loss=0.2231, simple_loss=0.2959, pruned_loss=0.07513, over 7979.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3014, pruned_loss=0.07302, over 1160634.87 frames. ], batch size: 21, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:02:19,376 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 16:02:28,582 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 16:02:43,773 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.666e+02 3.062e+02 4.026e+02 8.735e+02, threshold=6.124e+02, percent-clipped=4.0 +2023-02-06 16:02:44,420 INFO [train.py:901] (1/4) Epoch 15, batch 300, loss[loss=0.2653, simple_loss=0.3425, pruned_loss=0.094, over 8509.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3014, pruned_loss=0.0736, over 1258210.92 frames. ], batch size: 28, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:03:15,220 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 16:03:19,315 INFO [train.py:901] (1/4) Epoch 15, batch 350, loss[loss=0.1978, simple_loss=0.286, pruned_loss=0.05486, over 8474.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3024, pruned_loss=0.07353, over 1336931.70 frames. ], batch size: 25, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:03:52,038 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.415e+02 3.115e+02 3.728e+02 6.919e+02, threshold=6.229e+02, percent-clipped=2.0 +2023-02-06 16:03:52,738 INFO [train.py:901] (1/4) Epoch 15, batch 400, loss[loss=0.1882, simple_loss=0.2699, pruned_loss=0.05324, over 8090.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3023, pruned_loss=0.07363, over 1399321.81 frames. ], batch size: 21, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:04:17,470 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113596.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:04:28,817 INFO [train.py:901] (1/4) Epoch 15, batch 450, loss[loss=0.2138, simple_loss=0.2974, pruned_loss=0.06505, over 8516.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3008, pruned_loss=0.07218, over 1451830.91 frames. ], batch size: 26, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:04:30,184 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1542, 1.5032, 1.7044, 1.3576, 0.8740, 1.5017, 1.7574, 1.6388], + device='cuda:1'), covar=tensor([0.0465, 0.1159, 0.1645, 0.1358, 0.0626, 0.1431, 0.0666, 0.0604], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0155, 0.0102, 0.0162, 0.0115, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 16:04:43,277 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:04:56,096 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113654.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:05:01,032 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.616e+02 3.268e+02 4.141e+02 9.119e+02, threshold=6.536e+02, percent-clipped=2.0 +2023-02-06 16:05:01,751 INFO [train.py:901] (1/4) Epoch 15, batch 500, loss[loss=0.2001, simple_loss=0.2774, pruned_loss=0.06144, over 7534.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3001, pruned_loss=0.0713, over 1489879.07 frames. ], batch size: 18, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:05:02,554 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113664.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:05:12,315 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113679.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:05:35,423 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113711.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:05:36,632 INFO [train.py:901] (1/4) Epoch 15, batch 550, loss[loss=0.2007, simple_loss=0.2711, pruned_loss=0.06516, over 7686.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3013, pruned_loss=0.07242, over 1514816.94 frames. ], batch size: 18, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:06:09,825 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.516e+02 3.119e+02 4.209e+02 9.524e+02, threshold=6.239e+02, percent-clipped=4.0 +2023-02-06 16:06:10,535 INFO [train.py:901] (1/4) Epoch 15, batch 600, loss[loss=0.2081, simple_loss=0.2979, pruned_loss=0.05918, over 8293.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3017, pruned_loss=0.07244, over 1536157.71 frames. ], batch size: 23, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:06:17,859 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0010, 1.6059, 3.3661, 1.4181, 2.4345, 3.6289, 3.6683, 3.1768], + device='cuda:1'), covar=tensor([0.0996, 0.1525, 0.0299, 0.1970, 0.0877, 0.0229, 0.0470, 0.0524], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0308, 0.0273, 0.0301, 0.0288, 0.0247, 0.0377, 0.0299], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:06:24,211 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 16:06:44,318 INFO [train.py:901] (1/4) Epoch 15, batch 650, loss[loss=0.2155, simple_loss=0.2906, pruned_loss=0.07021, over 8568.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3017, pruned_loss=0.07253, over 1554845.79 frames. ], batch size: 31, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:07:01,571 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-06 16:07:19,223 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.270e+02 2.767e+02 3.649e+02 9.673e+02, threshold=5.535e+02, percent-clipped=4.0 +2023-02-06 16:07:19,890 INFO [train.py:901] (1/4) Epoch 15, batch 700, loss[loss=0.2371, simple_loss=0.3226, pruned_loss=0.07582, over 8331.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3006, pruned_loss=0.07157, over 1571460.62 frames. ], batch size: 25, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:07:28,851 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4068, 2.0180, 2.8468, 2.2688, 2.6332, 2.2537, 1.8983, 1.4390], + device='cuda:1'), covar=tensor([0.4414, 0.4363, 0.1524, 0.3072, 0.2262, 0.2570, 0.1902, 0.4698], + device='cuda:1'), in_proj_covar=tensor([0.0898, 0.0911, 0.0750, 0.0880, 0.0949, 0.0837, 0.0719, 0.0790], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 16:07:53,445 INFO [train.py:901] (1/4) Epoch 15, batch 750, loss[loss=0.212, simple_loss=0.2934, pruned_loss=0.06525, over 8457.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2995, pruned_loss=0.07166, over 1578964.21 frames. ], batch size: 25, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:08:11,154 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 16:08:20,445 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 16:08:28,023 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6319, 1.8805, 1.5790, 2.1866, 1.0562, 1.4324, 1.7509, 1.8916], + device='cuda:1'), covar=tensor([0.0813, 0.0664, 0.0969, 0.0508, 0.1117, 0.1367, 0.0714, 0.0699], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0203, 0.0250, 0.0211, 0.0211, 0.0248, 0.0252, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 16:08:29,183 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.267e+02 2.237e+02 2.791e+02 3.511e+02 6.350e+02, threshold=5.582e+02, percent-clipped=4.0 +2023-02-06 16:08:29,879 INFO [train.py:901] (1/4) Epoch 15, batch 800, loss[loss=0.221, simple_loss=0.3066, pruned_loss=0.06773, over 8105.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2984, pruned_loss=0.07099, over 1586145.65 frames. ], batch size: 23, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:08:32,837 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113967.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:08:40,773 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:08:50,090 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113992.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:09:01,972 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=114008.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:09:05,305 INFO [train.py:901] (1/4) Epoch 15, batch 850, loss[loss=0.2063, simple_loss=0.2896, pruned_loss=0.06146, over 7978.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2982, pruned_loss=0.07045, over 1595477.96 frames. ], batch size: 21, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:09:12,127 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9658, 2.4542, 2.8273, 1.3863, 2.9638, 1.7139, 1.5239, 1.9479], + device='cuda:1'), covar=tensor([0.0773, 0.0308, 0.0206, 0.0655, 0.0300, 0.0732, 0.0792, 0.0472], + device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0359, 0.0305, 0.0411, 0.0342, 0.0499, 0.0370, 0.0381], + device='cuda:1'), out_proj_covar=tensor([1.1633e-04, 9.6711e-05, 8.2020e-05, 1.1161e-04, 9.3094e-05, 1.4561e-04, + 1.0267e-04, 1.0432e-04], device='cuda:1') +2023-02-06 16:09:32,153 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0451, 1.4466, 1.6421, 1.3169, 0.9938, 1.4297, 1.6479, 1.3505], + device='cuda:1'), covar=tensor([0.0475, 0.1297, 0.1707, 0.1417, 0.0638, 0.1557, 0.0710, 0.0697], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0156, 0.0102, 0.0162, 0.0115, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 16:09:39,407 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.394e+02 2.826e+02 3.443e+02 6.296e+02, threshold=5.653e+02, percent-clipped=1.0 +2023-02-06 16:09:40,794 INFO [train.py:901] (1/4) Epoch 15, batch 900, loss[loss=0.2217, simple_loss=0.2997, pruned_loss=0.07191, over 8348.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2991, pruned_loss=0.07089, over 1603354.98 frames. ], batch size: 26, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:09:45,805 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 16:09:54,020 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3649, 1.9548, 2.8720, 2.2629, 2.7709, 2.1792, 1.8922, 1.4878], + device='cuda:1'), covar=tensor([0.4566, 0.4738, 0.1399, 0.2996, 0.2132, 0.2701, 0.1834, 0.4795], + device='cuda:1'), in_proj_covar=tensor([0.0890, 0.0903, 0.0742, 0.0872, 0.0938, 0.0828, 0.0711, 0.0781], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 16:09:58,192 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-02-06 16:10:02,633 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:10:15,168 INFO [train.py:901] (1/4) Epoch 15, batch 950, loss[loss=0.1903, simple_loss=0.2761, pruned_loss=0.05227, over 8138.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2997, pruned_loss=0.0708, over 1608459.20 frames. ], batch size: 22, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:10:21,910 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114123.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:10:36,822 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114145.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:10:38,929 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4896, 1.4268, 1.7804, 1.3384, 1.0745, 1.8296, 0.1353, 1.0346], + device='cuda:1'), covar=tensor([0.2210, 0.1458, 0.0524, 0.1299, 0.3418, 0.0458, 0.2946, 0.1902], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0179, 0.0111, 0.0217, 0.0259, 0.0114, 0.0165, 0.0175], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 16:10:39,443 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 16:10:49,210 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.449e+02 2.913e+02 3.851e+02 8.356e+02, threshold=5.826e+02, percent-clipped=3.0 +2023-02-06 16:10:49,928 INFO [train.py:901] (1/4) Epoch 15, batch 1000, loss[loss=0.2376, simple_loss=0.3161, pruned_loss=0.07954, over 8099.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3008, pruned_loss=0.07099, over 1612251.81 frames. ], batch size: 23, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:11:04,286 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4978, 2.7923, 2.0121, 2.2121, 2.2400, 1.5524, 2.0730, 2.1477], + device='cuda:1'), covar=tensor([0.1527, 0.0328, 0.1054, 0.0671, 0.0701, 0.1400, 0.1010, 0.0922], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0233, 0.0326, 0.0299, 0.0303, 0.0327, 0.0343, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:11:14,202 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 16:11:25,584 INFO [train.py:901] (1/4) Epoch 15, batch 1050, loss[loss=0.2232, simple_loss=0.2935, pruned_loss=0.07646, over 7802.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3001, pruned_loss=0.07096, over 1607845.76 frames. ], batch size: 19, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:11:25,599 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 16:11:57,612 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.714e+02 2.504e+02 3.058e+02 3.938e+02 1.189e+03, threshold=6.116e+02, percent-clipped=4.0 +2023-02-06 16:11:58,324 INFO [train.py:901] (1/4) Epoch 15, batch 1100, loss[loss=0.2019, simple_loss=0.2884, pruned_loss=0.05771, over 8474.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3004, pruned_loss=0.07149, over 1611592.46 frames. ], batch size: 49, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:12:11,164 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6747, 2.0378, 3.2687, 1.4196, 2.5451, 1.8873, 1.9150, 2.1881], + device='cuda:1'), covar=tensor([0.1856, 0.2497, 0.0890, 0.4672, 0.1703, 0.3380, 0.2091, 0.2506], + device='cuda:1'), in_proj_covar=tensor([0.0499, 0.0552, 0.0540, 0.0604, 0.0624, 0.0568, 0.0496, 0.0621], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:12:33,901 INFO [train.py:901] (1/4) Epoch 15, batch 1150, loss[loss=0.1872, simple_loss=0.2676, pruned_loss=0.05344, over 7537.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2976, pruned_loss=0.06997, over 1608966.77 frames. ], batch size: 18, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:12:38,620 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 16:12:59,528 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114350.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:13:07,347 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 2.463e+02 3.139e+02 3.955e+02 6.139e+02, threshold=6.277e+02, percent-clipped=1.0 +2023-02-06 16:13:07,972 INFO [train.py:901] (1/4) Epoch 15, batch 1200, loss[loss=0.191, simple_loss=0.2715, pruned_loss=0.05527, over 7539.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2974, pruned_loss=0.06971, over 1614360.34 frames. ], batch size: 18, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:13:16,131 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114375.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:13:18,735 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114379.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:13:30,065 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7160, 5.7888, 5.0654, 2.5543, 5.1288, 5.5381, 5.4434, 5.2599], + device='cuda:1'), covar=tensor([0.0542, 0.0376, 0.0892, 0.4060, 0.0641, 0.0706, 0.0898, 0.0565], + device='cuda:1'), in_proj_covar=tensor([0.0480, 0.0398, 0.0397, 0.0495, 0.0395, 0.0398, 0.0382, 0.0348], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:13:36,376 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114404.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:13:42,792 INFO [train.py:901] (1/4) Epoch 15, batch 1250, loss[loss=0.2588, simple_loss=0.3189, pruned_loss=0.09935, over 7653.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2984, pruned_loss=0.07019, over 1615606.86 frames. ], batch size: 19, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:14:16,862 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.591e+02 3.148e+02 4.129e+02 1.085e+03, threshold=6.295e+02, percent-clipped=6.0 +2023-02-06 16:14:17,475 INFO [train.py:901] (1/4) Epoch 15, batch 1300, loss[loss=0.2376, simple_loss=0.3205, pruned_loss=0.07732, over 8187.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2977, pruned_loss=0.07009, over 1613052.71 frames. ], batch size: 23, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:14:35,224 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=114489.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:14:51,246 INFO [train.py:901] (1/4) Epoch 15, batch 1350, loss[loss=0.1976, simple_loss=0.2654, pruned_loss=0.0649, over 7811.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.298, pruned_loss=0.0705, over 1614298.19 frames. ], batch size: 20, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:15:26,462 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.434e+02 2.903e+02 3.628e+02 5.826e+02, threshold=5.807e+02, percent-clipped=0.0 +2023-02-06 16:15:27,128 INFO [train.py:901] (1/4) Epoch 15, batch 1400, loss[loss=0.2261, simple_loss=0.3079, pruned_loss=0.07212, over 8513.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2988, pruned_loss=0.07124, over 1613204.70 frames. ], batch size: 26, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:15:54,552 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114604.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:16:00,687 INFO [train.py:901] (1/4) Epoch 15, batch 1450, loss[loss=0.1991, simple_loss=0.2836, pruned_loss=0.05723, over 8125.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2994, pruned_loss=0.07109, over 1613622.07 frames. ], batch size: 22, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:16:08,827 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 16:16:36,178 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.414e+02 3.068e+02 3.744e+02 6.619e+02, threshold=6.136e+02, percent-clipped=3.0 +2023-02-06 16:16:36,890 INFO [train.py:901] (1/4) Epoch 15, batch 1500, loss[loss=0.2198, simple_loss=0.2991, pruned_loss=0.07027, over 7920.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3005, pruned_loss=0.07155, over 1621004.65 frames. ], batch size: 20, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:17:11,514 INFO [train.py:901] (1/4) Epoch 15, batch 1550, loss[loss=0.2253, simple_loss=0.3095, pruned_loss=0.07056, over 8468.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3001, pruned_loss=0.07123, over 1625087.15 frames. ], batch size: 28, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:17:15,984 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 +2023-02-06 16:17:20,035 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4637, 2.2929, 4.1311, 1.3632, 2.8628, 2.0425, 1.7931, 2.4813], + device='cuda:1'), covar=tensor([0.2214, 0.2865, 0.0931, 0.4944, 0.1925, 0.3579, 0.2370, 0.3160], + device='cuda:1'), in_proj_covar=tensor([0.0500, 0.0551, 0.0539, 0.0602, 0.0623, 0.0566, 0.0494, 0.0624], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:17:26,187 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114734.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:17:45,705 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.278e+02 2.828e+02 3.736e+02 6.971e+02, threshold=5.655e+02, percent-clipped=1.0 +2023-02-06 16:17:46,444 INFO [train.py:901] (1/4) Epoch 15, batch 1600, loss[loss=0.2342, simple_loss=0.3091, pruned_loss=0.07969, over 8652.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2998, pruned_loss=0.0707, over 1627564.78 frames. ], batch size: 34, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:17:47,284 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6791, 2.2058, 3.1009, 1.9265, 1.7021, 3.1910, 0.7720, 2.0945], + device='cuda:1'), covar=tensor([0.1987, 0.1506, 0.0343, 0.2316, 0.3360, 0.0399, 0.2859, 0.1812], + device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0178, 0.0110, 0.0217, 0.0260, 0.0114, 0.0164, 0.0175], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 16:18:22,445 INFO [train.py:901] (1/4) Epoch 15, batch 1650, loss[loss=0.2261, simple_loss=0.3059, pruned_loss=0.07321, over 8516.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2994, pruned_loss=0.07044, over 1627143.82 frames. ], batch size: 28, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:18:55,128 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114860.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:18:56,281 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 2.429e+02 2.845e+02 3.384e+02 6.803e+02, threshold=5.691e+02, percent-clipped=1.0 +2023-02-06 16:18:56,976 INFO [train.py:901] (1/4) Epoch 15, batch 1700, loss[loss=0.2213, simple_loss=0.2951, pruned_loss=0.07372, over 8243.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2986, pruned_loss=0.07007, over 1624323.33 frames. ], batch size: 24, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:19:12,814 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114885.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:19:16,114 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114889.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:19:32,924 INFO [train.py:901] (1/4) Epoch 15, batch 1750, loss[loss=0.2353, simple_loss=0.3114, pruned_loss=0.07965, over 8479.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2999, pruned_loss=0.07063, over 1623335.42 frames. ], batch size: 25, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:19:45,240 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8316, 1.2816, 3.9685, 1.4379, 3.5088, 3.2491, 3.5630, 3.4450], + device='cuda:1'), covar=tensor([0.0633, 0.4492, 0.0610, 0.3806, 0.1181, 0.0969, 0.0647, 0.0739], + device='cuda:1'), in_proj_covar=tensor([0.0553, 0.0611, 0.0639, 0.0581, 0.0653, 0.0560, 0.0553, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 16:20:06,960 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.435e+02 3.025e+02 3.758e+02 7.531e+02, threshold=6.050e+02, percent-clipped=3.0 +2023-02-06 16:20:07,576 INFO [train.py:901] (1/4) Epoch 15, batch 1800, loss[loss=0.2751, simple_loss=0.3473, pruned_loss=0.1014, over 8455.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3014, pruned_loss=0.07115, over 1627171.72 frames. ], batch size: 27, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:20:43,789 INFO [train.py:901] (1/4) Epoch 15, batch 1850, loss[loss=0.1913, simple_loss=0.2683, pruned_loss=0.05717, over 6762.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3014, pruned_loss=0.07145, over 1623625.64 frames. ], batch size: 15, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:21:17,803 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.659e+02 3.189e+02 4.139e+02 1.250e+03, threshold=6.379e+02, percent-clipped=4.0 +2023-02-06 16:21:18,508 INFO [train.py:901] (1/4) Epoch 15, batch 1900, loss[loss=0.2021, simple_loss=0.2767, pruned_loss=0.06376, over 7777.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2995, pruned_loss=0.07049, over 1620620.54 frames. ], batch size: 19, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:21:28,839 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115078.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:21:46,754 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115104.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:21:50,093 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 16:21:53,560 INFO [train.py:901] (1/4) Epoch 15, batch 1950, loss[loss=0.1864, simple_loss=0.27, pruned_loss=0.05138, over 7926.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2984, pruned_loss=0.07023, over 1615500.70 frames. ], batch size: 20, lr: 5.13e-03, grad_scale: 32.0 +2023-02-06 16:22:04,500 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 16:22:11,847 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0646, 1.8186, 3.3986, 1.4365, 2.2921, 3.7105, 4.0059, 2.9646], + device='cuda:1'), covar=tensor([0.1134, 0.1706, 0.0382, 0.2268, 0.1200, 0.0296, 0.0456, 0.0759], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0305, 0.0270, 0.0297, 0.0285, 0.0245, 0.0373, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:22:23,206 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 16:22:28,430 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 2.421e+02 3.112e+02 3.916e+02 6.433e+02, threshold=6.224e+02, percent-clipped=1.0 +2023-02-06 16:22:29,136 INFO [train.py:901] (1/4) Epoch 15, batch 2000, loss[loss=0.2453, simple_loss=0.3225, pruned_loss=0.08407, over 8462.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2983, pruned_loss=0.07016, over 1613800.83 frames. ], batch size: 27, lr: 5.13e-03, grad_scale: 32.0 +2023-02-06 16:22:47,916 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-02-06 16:22:49,753 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115193.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:22:54,588 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3599, 2.4275, 1.7377, 1.9310, 2.0222, 1.4698, 1.8332, 1.8476], + device='cuda:1'), covar=tensor([0.1398, 0.0346, 0.1212, 0.0615, 0.0641, 0.1510, 0.0928, 0.0894], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0229, 0.0324, 0.0302, 0.0300, 0.0330, 0.0345, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:22:58,872 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9524, 2.0684, 1.8498, 2.5282, 1.2012, 1.5635, 1.9340, 2.1173], + device='cuda:1'), covar=tensor([0.0707, 0.0809, 0.0890, 0.0423, 0.1123, 0.1193, 0.0769, 0.0758], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0203, 0.0248, 0.0211, 0.0211, 0.0247, 0.0253, 0.0212], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 16:23:03,503 INFO [train.py:901] (1/4) Epoch 15, batch 2050, loss[loss=0.2316, simple_loss=0.3034, pruned_loss=0.07989, over 8472.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2976, pruned_loss=0.07042, over 1612839.25 frames. ], batch size: 25, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:23:18,039 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115233.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:23:36,042 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 16:23:39,482 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.382e+02 2.963e+02 3.753e+02 6.860e+02, threshold=5.925e+02, percent-clipped=2.0 +2023-02-06 16:23:39,503 INFO [train.py:901] (1/4) Epoch 15, batch 2100, loss[loss=0.2292, simple_loss=0.3169, pruned_loss=0.07076, over 8201.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2982, pruned_loss=0.07036, over 1613505.39 frames. ], batch size: 23, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:24:05,867 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9377, 1.3217, 6.0141, 2.2670, 5.4822, 5.0296, 5.5460, 5.3386], + device='cuda:1'), covar=tensor([0.0400, 0.4800, 0.0332, 0.3263, 0.0823, 0.0804, 0.0431, 0.0478], + device='cuda:1'), in_proj_covar=tensor([0.0543, 0.0598, 0.0628, 0.0568, 0.0643, 0.0552, 0.0542, 0.0609], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 16:24:13,859 INFO [train.py:901] (1/4) Epoch 15, batch 2150, loss[loss=0.2212, simple_loss=0.3091, pruned_loss=0.06664, over 8564.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2996, pruned_loss=0.07079, over 1618845.37 frames. ], batch size: 31, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:24:29,067 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5333, 1.9504, 2.0533, 1.1352, 2.1329, 1.4184, 0.5285, 1.8458], + device='cuda:1'), covar=tensor([0.0515, 0.0260, 0.0224, 0.0483, 0.0322, 0.0772, 0.0761, 0.0237], + device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0365, 0.0312, 0.0420, 0.0349, 0.0507, 0.0380, 0.0392], + device='cuda:1'), out_proj_covar=tensor([1.1842e-04, 9.8345e-05, 8.3735e-05, 1.1414e-04, 9.4909e-05, 1.4778e-04, + 1.0527e-04, 1.0719e-04], device='cuda:1') +2023-02-06 16:24:37,891 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115348.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:24:49,119 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.728e+02 2.532e+02 3.093e+02 4.065e+02 1.254e+03, threshold=6.185e+02, percent-clipped=7.0 +2023-02-06 16:24:49,139 INFO [train.py:901] (1/4) Epoch 15, batch 2200, loss[loss=0.2079, simple_loss=0.2804, pruned_loss=0.06772, over 8247.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2987, pruned_loss=0.07047, over 1620323.83 frames. ], batch size: 22, lr: 5.12e-03, grad_scale: 16.0 +2023-02-06 16:25:07,044 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:25:07,722 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:25:12,280 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0742, 1.7247, 3.0621, 1.4977, 2.2799, 3.2191, 3.3445, 2.7748], + device='cuda:1'), covar=tensor([0.0897, 0.1465, 0.0301, 0.1887, 0.0896, 0.0281, 0.0504, 0.0610], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0309, 0.0272, 0.0300, 0.0287, 0.0248, 0.0378, 0.0300], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:25:24,252 INFO [train.py:901] (1/4) Epoch 15, batch 2250, loss[loss=0.169, simple_loss=0.2434, pruned_loss=0.04727, over 7440.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2983, pruned_loss=0.07072, over 1614673.83 frames. ], batch size: 17, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:25:48,107 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115448.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:25:48,945 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115449.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:25:58,281 INFO [train.py:901] (1/4) Epoch 15, batch 2300, loss[loss=0.2162, simple_loss=0.2924, pruned_loss=0.07001, over 8244.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2988, pruned_loss=0.07093, over 1608338.88 frames. ], batch size: 22, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:25:58,960 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 2.502e+02 3.175e+02 3.927e+02 9.067e+02, threshold=6.350e+02, percent-clipped=5.0 +2023-02-06 16:26:04,775 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6048, 4.5951, 4.1090, 1.9978, 4.0804, 4.1663, 4.1431, 3.9231], + device='cuda:1'), covar=tensor([0.0714, 0.0513, 0.1086, 0.4617, 0.0907, 0.0998, 0.1190, 0.0823], + device='cuda:1'), in_proj_covar=tensor([0.0491, 0.0405, 0.0407, 0.0506, 0.0406, 0.0405, 0.0392, 0.0353], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:26:07,554 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115474.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:26:17,336 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6364, 1.4284, 1.6281, 1.2349, 0.8363, 1.3944, 1.4281, 1.2553], + device='cuda:1'), covar=tensor([0.0511, 0.1196, 0.1653, 0.1411, 0.0595, 0.1452, 0.0722, 0.0662], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0190, 0.0156, 0.0102, 0.0162, 0.0115, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 16:26:27,609 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6232, 2.0141, 3.2150, 1.3088, 2.2585, 2.0771, 1.6593, 2.2310], + device='cuda:1'), covar=tensor([0.1785, 0.2441, 0.0700, 0.4190, 0.1947, 0.2836, 0.2022, 0.2312], + device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0548, 0.0539, 0.0601, 0.0621, 0.0564, 0.0493, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:26:34,675 INFO [train.py:901] (1/4) Epoch 15, batch 2350, loss[loss=0.1981, simple_loss=0.2765, pruned_loss=0.05987, over 7544.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.299, pruned_loss=0.07085, over 1610664.95 frames. ], batch size: 18, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:27:09,321 INFO [train.py:901] (1/4) Epoch 15, batch 2400, loss[loss=0.2094, simple_loss=0.3053, pruned_loss=0.05677, over 8504.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2988, pruned_loss=0.07117, over 1608831.26 frames. ], batch size: 26, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:27:09,509 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115563.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:27:10,002 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.542e+02 3.047e+02 3.524e+02 9.073e+02, threshold=6.095e+02, percent-clipped=1.0 +2023-02-06 16:27:39,713 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115604.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:27:45,574 INFO [train.py:901] (1/4) Epoch 15, batch 2450, loss[loss=0.2072, simple_loss=0.2923, pruned_loss=0.06108, over 8512.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2986, pruned_loss=0.07022, over 1612547.83 frames. ], batch size: 28, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:27:56,554 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115629.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:28:19,908 INFO [train.py:901] (1/4) Epoch 15, batch 2500, loss[loss=0.261, simple_loss=0.3258, pruned_loss=0.09816, over 8425.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.297, pruned_loss=0.06988, over 1611588.49 frames. ], batch size: 27, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:28:20,562 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.367e+02 2.686e+02 3.697e+02 9.165e+02, threshold=5.372e+02, percent-clipped=5.0 +2023-02-06 16:28:40,516 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9915, 1.4503, 1.5909, 1.2447, 0.8508, 1.3997, 1.7072, 1.7057], + device='cuda:1'), covar=tensor([0.0509, 0.1329, 0.1798, 0.1503, 0.0629, 0.1586, 0.0728, 0.0603], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0157, 0.0102, 0.0162, 0.0115, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 16:28:55,237 INFO [train.py:901] (1/4) Epoch 15, batch 2550, loss[loss=0.1884, simple_loss=0.2573, pruned_loss=0.0598, over 7710.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2978, pruned_loss=0.07035, over 1610836.43 frames. ], batch size: 18, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:29:05,128 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 +2023-02-06 16:29:08,941 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115732.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:29:09,605 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:29:14,879 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7055, 1.7552, 2.2459, 1.5860, 1.0262, 2.2507, 0.2753, 1.3377], + device='cuda:1'), covar=tensor([0.2175, 0.1431, 0.0410, 0.1707, 0.4068, 0.0400, 0.3052, 0.1806], + device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0176, 0.0110, 0.0214, 0.0258, 0.0115, 0.0163, 0.0174], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 16:29:30,403 INFO [train.py:901] (1/4) Epoch 15, batch 2600, loss[loss=0.203, simple_loss=0.2858, pruned_loss=0.06017, over 8296.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.298, pruned_loss=0.07052, over 1611858.31 frames. ], batch size: 23, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:29:31,074 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.554e+02 2.427e+02 3.148e+02 3.839e+02 8.607e+02, threshold=6.296e+02, percent-clipped=3.0 +2023-02-06 16:30:04,163 INFO [train.py:901] (1/4) Epoch 15, batch 2650, loss[loss=0.1856, simple_loss=0.2612, pruned_loss=0.05498, over 7441.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2984, pruned_loss=0.07092, over 1607568.21 frames. ], batch size: 17, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:30:08,492 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115819.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:30:27,410 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115844.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:30:29,376 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115847.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:30:30,057 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:30:39,733 INFO [train.py:901] (1/4) Epoch 15, batch 2700, loss[loss=0.2582, simple_loss=0.3222, pruned_loss=0.09711, over 5559.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2978, pruned_loss=0.07032, over 1605496.91 frames. ], batch size: 12, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:30:40,394 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.337e+02 2.718e+02 3.606e+02 6.832e+02, threshold=5.436e+02, percent-clipped=3.0 +2023-02-06 16:31:12,012 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115910.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 16:31:13,854 INFO [train.py:901] (1/4) Epoch 15, batch 2750, loss[loss=0.2008, simple_loss=0.2869, pruned_loss=0.05736, over 8715.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2979, pruned_loss=0.06998, over 1611874.31 frames. ], batch size: 39, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:31:49,502 INFO [train.py:901] (1/4) Epoch 15, batch 2800, loss[loss=0.2128, simple_loss=0.283, pruned_loss=0.07133, over 8222.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2985, pruned_loss=0.07083, over 1606863.70 frames. ], batch size: 22, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:31:50,149 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.517e+02 2.986e+02 3.677e+02 9.071e+02, threshold=5.972e+02, percent-clipped=5.0 +2023-02-06 16:31:53,143 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5805, 2.0173, 3.3448, 1.4000, 2.3209, 2.0504, 1.5931, 2.4133], + device='cuda:1'), covar=tensor([0.1871, 0.2306, 0.0833, 0.4204, 0.1831, 0.2894, 0.2161, 0.2106], + device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0551, 0.0541, 0.0601, 0.0624, 0.0567, 0.0496, 0.0619], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:32:16,425 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.05 vs. limit=5.0 +2023-02-06 16:32:24,935 INFO [train.py:901] (1/4) Epoch 15, batch 2850, loss[loss=0.2492, simple_loss=0.3322, pruned_loss=0.08312, over 8733.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2989, pruned_loss=0.07098, over 1609873.23 frames. ], batch size: 30, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:32:38,104 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:00,842 INFO [train.py:901] (1/4) Epoch 15, batch 2900, loss[loss=0.183, simple_loss=0.2632, pruned_loss=0.05141, over 7523.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2985, pruned_loss=0.07099, over 1609541.56 frames. ], batch size: 18, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:33:01,416 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.452e+02 2.959e+02 3.782e+02 6.842e+02, threshold=5.917e+02, percent-clipped=3.0 +2023-02-06 16:33:29,123 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116103.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:29,801 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116104.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:35,325 INFO [train.py:901] (1/4) Epoch 15, batch 2950, loss[loss=0.1967, simple_loss=0.2645, pruned_loss=0.06442, over 7441.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2987, pruned_loss=0.07082, over 1612852.30 frames. ], batch size: 17, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:33:36,699 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 16:33:42,121 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116123.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:44,354 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 +2023-02-06 16:33:45,593 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:46,297 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116129.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:49,699 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:34:01,862 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1455, 4.0346, 2.5531, 2.7167, 2.8525, 2.0448, 2.6072, 2.9566], + device='cuda:1'), covar=tensor([0.1766, 0.0343, 0.0983, 0.0850, 0.0765, 0.1373, 0.1198, 0.1146], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0230, 0.0325, 0.0302, 0.0300, 0.0330, 0.0344, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:34:08,982 INFO [train.py:901] (1/4) Epoch 15, batch 3000, loss[loss=0.208, simple_loss=0.2849, pruned_loss=0.06555, over 7914.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2989, pruned_loss=0.07127, over 1611721.29 frames. ], batch size: 20, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:34:08,982 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 16:34:21,680 INFO [train.py:935] (1/4) Epoch 15, validation: loss=0.1808, simple_loss=0.2809, pruned_loss=0.04034, over 944034.00 frames. +2023-02-06 16:34:21,681 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 16:34:22,362 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.534e+02 3.127e+02 3.845e+02 7.463e+02, threshold=6.253e+02, percent-clipped=8.0 +2023-02-06 16:34:29,498 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 16:34:55,445 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6700, 1.9508, 3.1544, 1.3597, 2.3870, 2.0106, 1.6212, 2.3312], + device='cuda:1'), covar=tensor([0.1685, 0.2200, 0.0843, 0.3988, 0.1551, 0.2754, 0.2031, 0.2072], + device='cuda:1'), in_proj_covar=tensor([0.0494, 0.0550, 0.0539, 0.0598, 0.0621, 0.0565, 0.0493, 0.0616], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:34:57,896 INFO [train.py:901] (1/4) Epoch 15, batch 3050, loss[loss=0.1748, simple_loss=0.2554, pruned_loss=0.0471, over 7540.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.298, pruned_loss=0.07089, over 1608058.47 frames. ], batch size: 18, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:35:09,549 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5226, 1.9689, 3.1947, 1.3193, 2.3394, 1.9374, 1.5898, 2.2987], + device='cuda:1'), covar=tensor([0.1805, 0.2396, 0.0770, 0.4159, 0.1613, 0.3008, 0.2033, 0.2234], + device='cuda:1'), in_proj_covar=tensor([0.0497, 0.0553, 0.0542, 0.0600, 0.0624, 0.0567, 0.0496, 0.0619], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:35:15,120 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 16:35:26,167 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116254.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 16:35:31,951 INFO [train.py:901] (1/4) Epoch 15, batch 3100, loss[loss=0.2484, simple_loss=0.329, pruned_loss=0.08387, over 8787.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2986, pruned_loss=0.07098, over 1612738.06 frames. ], batch size: 40, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:35:32,579 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.573e+02 3.095e+02 3.865e+02 1.142e+03, threshold=6.190e+02, percent-clipped=3.0 +2023-02-06 16:35:37,003 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 +2023-02-06 16:35:46,421 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4659, 1.9107, 4.5060, 2.0459, 2.3007, 5.1121, 5.0819, 4.4746], + device='cuda:1'), covar=tensor([0.1034, 0.1548, 0.0262, 0.1775, 0.1219, 0.0171, 0.0393, 0.0515], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0304, 0.0270, 0.0297, 0.0286, 0.0248, 0.0376, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:35:59,165 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.90 vs. limit=5.0 +2023-02-06 16:36:06,932 INFO [train.py:901] (1/4) Epoch 15, batch 3150, loss[loss=0.2231, simple_loss=0.311, pruned_loss=0.06763, over 8185.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2975, pruned_loss=0.07036, over 1606687.87 frames. ], batch size: 23, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:36:15,222 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1973, 1.1454, 1.2928, 1.0765, 0.9552, 1.2923, 0.0224, 0.8623], + device='cuda:1'), covar=tensor([0.2035, 0.1464, 0.0546, 0.1080, 0.3218, 0.0575, 0.2950, 0.1699], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0175, 0.0109, 0.0212, 0.0254, 0.0113, 0.0160, 0.0174], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 16:36:15,822 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2946, 2.2432, 2.1076, 1.2612, 2.0712, 2.0819, 2.0895, 1.9343], + device='cuda:1'), covar=tensor([0.1072, 0.0903, 0.1069, 0.3543, 0.0953, 0.1222, 0.1346, 0.1081], + device='cuda:1'), in_proj_covar=tensor([0.0483, 0.0398, 0.0402, 0.0497, 0.0394, 0.0396, 0.0387, 0.0347], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:36:27,180 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116341.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:36:30,072 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1946, 1.5325, 1.5295, 1.3683, 0.8781, 1.3422, 1.7394, 1.9387], + device='cuda:1'), covar=tensor([0.0469, 0.1274, 0.1745, 0.1436, 0.0625, 0.1551, 0.0715, 0.0522], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0156, 0.0101, 0.0162, 0.0115, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 16:36:41,971 INFO [train.py:901] (1/4) Epoch 15, batch 3200, loss[loss=0.23, simple_loss=0.3099, pruned_loss=0.07509, over 8097.00 frames. ], tot_loss[loss=0.221, simple_loss=0.299, pruned_loss=0.07147, over 1606494.63 frames. ], batch size: 23, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:36:43,343 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.524e+02 3.304e+02 3.942e+02 1.206e+03, threshold=6.608e+02, percent-clipped=2.0 +2023-02-06 16:36:46,721 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116369.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 16:36:51,222 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:36:53,705 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 16:37:06,193 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6100, 2.9906, 2.4804, 4.0902, 1.7512, 2.3072, 2.2569, 3.0796], + device='cuda:1'), covar=tensor([0.0662, 0.0781, 0.0861, 0.0226, 0.1146, 0.1273, 0.1131, 0.0811], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0203, 0.0250, 0.0212, 0.0211, 0.0250, 0.0257, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 16:37:12,413 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9824, 1.7804, 3.2975, 1.3937, 2.1652, 3.5780, 3.7020, 3.0843], + device='cuda:1'), covar=tensor([0.1023, 0.1449, 0.0308, 0.2009, 0.1024, 0.0238, 0.0508, 0.0568], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0306, 0.0272, 0.0299, 0.0288, 0.0248, 0.0377, 0.0299], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:37:16,499 INFO [train.py:901] (1/4) Epoch 15, batch 3250, loss[loss=0.2157, simple_loss=0.3005, pruned_loss=0.0655, over 8133.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2992, pruned_loss=0.07131, over 1611106.77 frames. ], batch size: 22, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:37:52,484 INFO [train.py:901] (1/4) Epoch 15, batch 3300, loss[loss=0.2299, simple_loss=0.3032, pruned_loss=0.07828, over 7932.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2992, pruned_loss=0.07113, over 1612915.87 frames. ], batch size: 20, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:37:53,148 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.388e+02 2.875e+02 3.716e+02 9.209e+02, threshold=5.750e+02, percent-clipped=3.0 +2023-02-06 16:37:53,292 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116464.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:37:55,221 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116467.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:38:02,543 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:38:12,009 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116491.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:38:26,442 INFO [train.py:901] (1/4) Epoch 15, batch 3350, loss[loss=0.2236, simple_loss=0.3097, pruned_loss=0.06876, over 8183.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.299, pruned_loss=0.07111, over 1613010.87 frames. ], batch size: 23, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:38:33,231 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116523.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:39:02,049 INFO [train.py:901] (1/4) Epoch 15, batch 3400, loss[loss=0.1994, simple_loss=0.2705, pruned_loss=0.06418, over 7703.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2979, pruned_loss=0.07042, over 1610730.75 frames. ], batch size: 18, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:39:02,716 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.566e+02 3.149e+02 4.104e+02 8.501e+02, threshold=6.298e+02, percent-clipped=7.0 +2023-02-06 16:39:14,872 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116582.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:39:22,204 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116593.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:39:36,209 INFO [train.py:901] (1/4) Epoch 15, batch 3450, loss[loss=0.2464, simple_loss=0.3126, pruned_loss=0.09009, over 8472.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.299, pruned_loss=0.07066, over 1614383.91 frames. ], batch size: 27, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:39:44,410 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116625.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 16:39:51,719 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:40:01,117 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116650.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 16:40:03,771 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0742, 1.2464, 1.1832, 0.6207, 1.2209, 0.9716, 0.1267, 1.2272], + device='cuda:1'), covar=tensor([0.0355, 0.0286, 0.0269, 0.0439, 0.0320, 0.0779, 0.0655, 0.0245], + device='cuda:1'), in_proj_covar=tensor([0.0424, 0.0363, 0.0311, 0.0419, 0.0351, 0.0507, 0.0375, 0.0385], + device='cuda:1'), out_proj_covar=tensor([1.1716e-04, 9.7643e-05, 8.3560e-05, 1.1352e-04, 9.5440e-05, 1.4759e-04, + 1.0361e-04, 1.0489e-04], device='cuda:1') +2023-02-06 16:40:10,038 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.07 vs. limit=5.0 +2023-02-06 16:40:10,175 INFO [train.py:901] (1/4) Epoch 15, batch 3500, loss[loss=0.2018, simple_loss=0.2967, pruned_loss=0.05348, over 7966.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3002, pruned_loss=0.07097, over 1620113.90 frames. ], batch size: 21, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:40:10,857 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.712e+02 2.398e+02 2.936e+02 3.935e+02 9.560e+02, threshold=5.871e+02, percent-clipped=3.0 +2023-02-06 16:40:13,168 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5818, 1.9800, 3.3212, 1.3536, 2.5446, 1.9774, 1.6976, 2.3496], + device='cuda:1'), covar=tensor([0.1942, 0.2500, 0.0820, 0.4502, 0.1765, 0.3148, 0.2151, 0.2419], + device='cuda:1'), in_proj_covar=tensor([0.0503, 0.0557, 0.0547, 0.0609, 0.0628, 0.0571, 0.0502, 0.0627], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:40:26,407 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116685.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:40:35,640 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 16:40:44,877 INFO [train.py:901] (1/4) Epoch 15, batch 3550, loss[loss=0.2211, simple_loss=0.3046, pruned_loss=0.06878, over 8327.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2996, pruned_loss=0.07084, over 1618790.56 frames. ], batch size: 25, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:40:50,941 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116722.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:06,498 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9870, 1.4599, 4.4746, 2.0760, 2.3707, 5.0574, 5.1163, 4.3893], + device='cuda:1'), covar=tensor([0.1179, 0.1771, 0.0249, 0.1881, 0.1160, 0.0180, 0.0355, 0.0512], + device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0308, 0.0272, 0.0304, 0.0291, 0.0251, 0.0380, 0.0302], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:41:08,579 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116747.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:17,775 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 16:41:19,406 INFO [train.py:901] (1/4) Epoch 15, batch 3600, loss[loss=0.2219, simple_loss=0.2984, pruned_loss=0.07273, over 8627.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3011, pruned_loss=0.07162, over 1619970.97 frames. ], batch size: 39, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:41:20,116 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.627e+02 3.005e+02 3.918e+02 8.490e+02, threshold=6.010e+02, percent-clipped=4.0 +2023-02-06 16:41:25,803 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:45,870 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6733, 1.5909, 3.1920, 1.1453, 2.3017, 3.4961, 3.7822, 2.6344], + device='cuda:1'), covar=tensor([0.1607, 0.1934, 0.0494, 0.2873, 0.1146, 0.0454, 0.0627, 0.1203], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0307, 0.0273, 0.0303, 0.0291, 0.0251, 0.0379, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:41:47,429 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116800.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:52,897 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116808.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:56,181 INFO [train.py:901] (1/4) Epoch 15, batch 3650, loss[loss=0.2271, simple_loss=0.2976, pruned_loss=0.07827, over 7702.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.299, pruned_loss=0.0706, over 1616069.71 frames. ], batch size: 18, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:41:56,581 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 16:42:00,909 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:42:13,591 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116838.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:42:21,044 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116849.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:42:30,282 INFO [train.py:901] (1/4) Epoch 15, batch 3700, loss[loss=0.2275, simple_loss=0.3034, pruned_loss=0.07578, over 8105.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.299, pruned_loss=0.07096, over 1612518.45 frames. ], batch size: 23, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:42:30,497 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:42:30,964 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.301e+02 2.797e+02 3.414e+02 8.630e+02, threshold=5.595e+02, percent-clipped=3.0 +2023-02-06 16:42:33,144 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116867.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:42:36,596 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 16:42:38,095 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116874.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:43:06,649 INFO [train.py:901] (1/4) Epoch 15, batch 3750, loss[loss=0.1609, simple_loss=0.2354, pruned_loss=0.04317, over 7412.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2977, pruned_loss=0.07031, over 1610282.09 frames. ], batch size: 17, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:43:13,675 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116923.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:43:14,068 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-06 16:43:40,818 INFO [train.py:901] (1/4) Epoch 15, batch 3800, loss[loss=0.2161, simple_loss=0.2848, pruned_loss=0.07365, over 7252.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2972, pruned_loss=0.07018, over 1608910.27 frames. ], batch size: 16, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:43:41,468 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.512e+02 2.989e+02 3.697e+02 7.171e+02, threshold=5.977e+02, percent-clipped=7.0 +2023-02-06 16:43:52,422 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116980.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:43:53,919 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116982.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:44:15,608 INFO [train.py:901] (1/4) Epoch 15, batch 3850, loss[loss=0.2475, simple_loss=0.3164, pruned_loss=0.08931, over 8624.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2972, pruned_loss=0.06967, over 1612079.46 frames. ], batch size: 49, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:44:42,564 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 16:44:46,214 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117056.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:44:46,552 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.72 vs. limit=5.0 +2023-02-06 16:44:50,972 INFO [train.py:901] (1/4) Epoch 15, batch 3900, loss[loss=0.1791, simple_loss=0.262, pruned_loss=0.04816, over 8135.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2984, pruned_loss=0.06993, over 1618358.94 frames. ], batch size: 22, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:44:51,622 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 2.428e+02 3.027e+02 3.797e+02 6.654e+02, threshold=6.053e+02, percent-clipped=2.0 +2023-02-06 16:44:53,045 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:45:03,688 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117081.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:45:12,955 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117095.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:45:17,722 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5068, 2.7120, 1.9365, 2.2300, 2.2314, 1.6163, 1.9414, 2.1430], + device='cuda:1'), covar=tensor([0.1484, 0.0368, 0.1129, 0.0663, 0.0717, 0.1415, 0.1115, 0.0923], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0233, 0.0327, 0.0306, 0.0303, 0.0333, 0.0349, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 16:45:24,920 INFO [train.py:901] (1/4) Epoch 15, batch 3950, loss[loss=0.1879, simple_loss=0.2637, pruned_loss=0.05609, over 7555.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2971, pruned_loss=0.06872, over 1617339.97 frames. ], batch size: 18, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:45:34,223 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 16:46:01,091 INFO [train.py:901] (1/4) Epoch 15, batch 4000, loss[loss=0.2364, simple_loss=0.3233, pruned_loss=0.07477, over 8322.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2996, pruned_loss=0.07036, over 1616541.17 frames. ], batch size: 25, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:46:01,784 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.728e+02 2.473e+02 2.992e+02 3.534e+02 5.115e+02, threshold=5.984e+02, percent-clipped=0.0 +2023-02-06 16:46:01,906 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117164.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:46:12,549 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117179.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:46:13,870 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117181.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:46:15,175 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7425, 1.5404, 3.8916, 1.5459, 3.3647, 3.2044, 3.5345, 3.3895], + device='cuda:1'), covar=tensor([0.0730, 0.4229, 0.0797, 0.3819, 0.1440, 0.1118, 0.0678, 0.0835], + device='cuda:1'), in_proj_covar=tensor([0.0554, 0.0602, 0.0633, 0.0571, 0.0654, 0.0556, 0.0548, 0.0613], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 16:46:17,463 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-06 16:46:29,841 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117204.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:46:35,577 INFO [train.py:901] (1/4) Epoch 15, batch 4050, loss[loss=0.219, simple_loss=0.2875, pruned_loss=0.07521, over 8238.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2989, pruned_loss=0.07029, over 1613358.61 frames. ], batch size: 22, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:46:53,310 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117238.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:46:53,896 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117239.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:47:11,623 INFO [train.py:901] (1/4) Epoch 15, batch 4100, loss[loss=0.2276, simple_loss=0.307, pruned_loss=0.07407, over 8466.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.299, pruned_loss=0.07027, over 1611558.79 frames. ], batch size: 25, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:47:11,821 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117263.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:47:12,282 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.506e+02 3.096e+02 3.742e+02 9.544e+02, threshold=6.191e+02, percent-clipped=4.0 +2023-02-06 16:47:22,916 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117279.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:47:46,643 INFO [train.py:901] (1/4) Epoch 15, batch 4150, loss[loss=0.2348, simple_loss=0.316, pruned_loss=0.07683, over 8475.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2998, pruned_loss=0.07081, over 1609563.90 frames. ], batch size: 50, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:48:10,091 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:48:12,975 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117351.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:48:21,502 INFO [train.py:901] (1/4) Epoch 15, batch 4200, loss[loss=0.2061, simple_loss=0.2968, pruned_loss=0.05765, over 8284.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2986, pruned_loss=0.07044, over 1606795.39 frames. ], batch size: 23, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:48:22,820 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.404e+02 2.907e+02 3.383e+02 1.073e+03, threshold=5.814e+02, percent-clipped=1.0 +2023-02-06 16:48:31,972 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:48:40,573 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 16:48:48,359 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8669, 1.6394, 5.9938, 2.3516, 5.4031, 5.0655, 5.5290, 5.4032], + device='cuda:1'), covar=tensor([0.0456, 0.4761, 0.0395, 0.3475, 0.0890, 0.0797, 0.0471, 0.0480], + device='cuda:1'), in_proj_covar=tensor([0.0550, 0.0605, 0.0628, 0.0571, 0.0650, 0.0552, 0.0546, 0.0610], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 16:48:57,042 INFO [train.py:901] (1/4) Epoch 15, batch 4250, loss[loss=0.2143, simple_loss=0.3017, pruned_loss=0.06345, over 8022.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2996, pruned_loss=0.071, over 1610892.39 frames. ], batch size: 22, lr: 5.08e-03, grad_scale: 16.0 +2023-02-06 16:49:03,725 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 16:49:14,102 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117437.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:49:30,900 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117462.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:49:31,426 INFO [train.py:901] (1/4) Epoch 15, batch 4300, loss[loss=0.2597, simple_loss=0.3415, pruned_loss=0.08891, over 8316.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3004, pruned_loss=0.07164, over 1615324.68 frames. ], batch size: 25, lr: 5.08e-03, grad_scale: 16.0 +2023-02-06 16:49:32,090 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.479e+02 3.115e+02 3.892e+02 7.815e+02, threshold=6.229e+02, percent-clipped=5.0 +2023-02-06 16:50:05,920 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 16:50:07,599 INFO [train.py:901] (1/4) Epoch 15, batch 4350, loss[loss=0.1972, simple_loss=0.2816, pruned_loss=0.05644, over 8082.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3006, pruned_loss=0.07128, over 1619034.84 frames. ], batch size: 21, lr: 5.08e-03, grad_scale: 16.0 +2023-02-06 16:50:23,045 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117535.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:50:36,326 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 16:50:40,556 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117560.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:50:42,361 INFO [train.py:901] (1/4) Epoch 15, batch 4400, loss[loss=0.2146, simple_loss=0.2915, pruned_loss=0.06886, over 7257.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3008, pruned_loss=0.07167, over 1611700.37 frames. ], batch size: 16, lr: 5.08e-03, grad_scale: 16.0 +2023-02-06 16:50:43,035 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.383e+02 3.124e+02 3.901e+02 9.506e+02, threshold=6.248e+02, percent-clipped=7.0 +2023-02-06 16:50:55,789 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:51:17,967 INFO [train.py:901] (1/4) Epoch 15, batch 4450, loss[loss=0.1883, simple_loss=0.2749, pruned_loss=0.05085, over 7984.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2998, pruned_loss=0.07113, over 1611792.40 frames. ], batch size: 21, lr: 5.07e-03, grad_scale: 16.0 +2023-02-06 16:51:17,986 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 16:51:52,096 INFO [train.py:901] (1/4) Epoch 15, batch 4500, loss[loss=0.2171, simple_loss=0.3033, pruned_loss=0.06547, over 8613.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3012, pruned_loss=0.0727, over 1611653.85 frames. ], batch size: 34, lr: 5.07e-03, grad_scale: 16.0 +2023-02-06 16:51:52,740 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.719e+02 2.480e+02 2.963e+02 4.043e+02 1.091e+03, threshold=5.927e+02, percent-clipped=5.0 +2023-02-06 16:52:11,217 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:52:11,859 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 16:52:16,201 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:52:27,012 INFO [train.py:901] (1/4) Epoch 15, batch 4550, loss[loss=0.2087, simple_loss=0.3025, pruned_loss=0.05739, over 8499.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3001, pruned_loss=0.07168, over 1609434.70 frames. ], batch size: 29, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:52:32,133 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 16:52:42,013 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0951, 4.0021, 3.6188, 1.9104, 3.5374, 3.6551, 3.7098, 3.4613], + device='cuda:1'), covar=tensor([0.0849, 0.0705, 0.1077, 0.5183, 0.0917, 0.1159, 0.1343, 0.0942], + device='cuda:1'), in_proj_covar=tensor([0.0485, 0.0403, 0.0405, 0.0503, 0.0396, 0.0405, 0.0387, 0.0350], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:53:02,114 INFO [train.py:901] (1/4) Epoch 15, batch 4600, loss[loss=0.236, simple_loss=0.3207, pruned_loss=0.07568, over 8252.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2996, pruned_loss=0.07169, over 1610386.17 frames. ], batch size: 24, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:53:03,483 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.311e+02 2.848e+02 3.671e+02 5.923e+02, threshold=5.697e+02, percent-clipped=0.0 +2023-02-06 16:53:26,848 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6431, 1.8857, 2.0251, 1.2095, 2.1277, 1.4019, 0.5072, 1.8609], + device='cuda:1'), covar=tensor([0.0467, 0.0290, 0.0244, 0.0469, 0.0325, 0.0791, 0.0694, 0.0221], + device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0360, 0.0311, 0.0415, 0.0345, 0.0506, 0.0375, 0.0382], + device='cuda:1'), out_proj_covar=tensor([1.1561e-04, 9.6923e-05, 8.3245e-05, 1.1243e-04, 9.3747e-05, 1.4756e-04, + 1.0351e-04, 1.0386e-04], device='cuda:1') +2023-02-06 16:53:31,663 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117806.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:53:36,046 INFO [train.py:901] (1/4) Epoch 15, batch 4650, loss[loss=0.2121, simple_loss=0.2821, pruned_loss=0.07105, over 7443.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2985, pruned_loss=0.07122, over 1611076.60 frames. ], batch size: 17, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:53:58,334 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1522, 4.0863, 3.7176, 1.7991, 3.6179, 3.7434, 3.7589, 3.5029], + device='cuda:1'), covar=tensor([0.0854, 0.0652, 0.1195, 0.5196, 0.0955, 0.0970, 0.1337, 0.0864], + device='cuda:1'), in_proj_covar=tensor([0.0490, 0.0407, 0.0411, 0.0510, 0.0402, 0.0409, 0.0391, 0.0353], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:54:11,643 INFO [train.py:901] (1/4) Epoch 15, batch 4700, loss[loss=0.2219, simple_loss=0.3036, pruned_loss=0.07012, over 8240.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2979, pruned_loss=0.07075, over 1610243.35 frames. ], batch size: 24, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:54:11,861 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5151, 1.9031, 3.3094, 1.3653, 2.4614, 2.0160, 1.5628, 2.4178], + device='cuda:1'), covar=tensor([0.1835, 0.2261, 0.0650, 0.3792, 0.1631, 0.2726, 0.2042, 0.2102], + device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0549, 0.0534, 0.0597, 0.0620, 0.0563, 0.0491, 0.0616], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:54:12,891 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.509e+02 3.109e+02 4.231e+02 8.316e+02, threshold=6.217e+02, percent-clipped=12.0 +2023-02-06 16:54:21,295 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 16:54:46,558 INFO [train.py:901] (1/4) Epoch 15, batch 4750, loss[loss=0.2121, simple_loss=0.2919, pruned_loss=0.06613, over 8240.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2988, pruned_loss=0.07147, over 1608870.38 frames. ], batch size: 24, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:55:11,978 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 16:55:15,302 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 16:55:16,047 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117954.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:55:22,513 INFO [train.py:901] (1/4) Epoch 15, batch 4800, loss[loss=0.2593, simple_loss=0.3228, pruned_loss=0.09795, over 7923.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2993, pruned_loss=0.07151, over 1607548.33 frames. ], batch size: 20, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:55:23,940 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.482e+02 3.121e+02 4.555e+02 1.692e+03, threshold=6.242e+02, percent-clipped=8.0 +2023-02-06 16:55:33,842 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:55:57,731 INFO [train.py:901] (1/4) Epoch 15, batch 4850, loss[loss=0.2026, simple_loss=0.2998, pruned_loss=0.05265, over 8247.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2996, pruned_loss=0.07161, over 1608927.11 frames. ], batch size: 24, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:56:07,039 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 16:56:29,206 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118058.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:56:29,474 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 16:56:31,981 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:56:32,444 INFO [train.py:901] (1/4) Epoch 15, batch 4900, loss[loss=0.2052, simple_loss=0.284, pruned_loss=0.06318, over 7424.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2986, pruned_loss=0.07089, over 1611293.73 frames. ], batch size: 17, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:56:33,723 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.453e+02 2.951e+02 3.688e+02 9.605e+02, threshold=5.903e+02, percent-clipped=5.0 +2023-02-06 16:56:50,260 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:56:56,421 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8876, 1.6068, 2.0412, 1.7672, 1.9452, 1.8630, 1.6348, 0.7563], + device='cuda:1'), covar=tensor([0.4928, 0.4161, 0.1560, 0.3055, 0.2133, 0.2586, 0.1871, 0.4482], + device='cuda:1'), in_proj_covar=tensor([0.0908, 0.0914, 0.0752, 0.0882, 0.0950, 0.0838, 0.0717, 0.0793], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 16:57:07,609 INFO [train.py:901] (1/4) Epoch 15, batch 4950, loss[loss=0.2035, simple_loss=0.277, pruned_loss=0.06497, over 8243.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2984, pruned_loss=0.07072, over 1609305.36 frames. ], batch size: 22, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:57:33,427 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2422, 4.1733, 3.7615, 2.2043, 3.6858, 3.8488, 3.9000, 3.5307], + device='cuda:1'), covar=tensor([0.0856, 0.0696, 0.1081, 0.4317, 0.0919, 0.1028, 0.1259, 0.0914], + device='cuda:1'), in_proj_covar=tensor([0.0484, 0.0405, 0.0408, 0.0507, 0.0399, 0.0407, 0.0388, 0.0351], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 16:57:42,121 INFO [train.py:901] (1/4) Epoch 15, batch 5000, loss[loss=0.2171, simple_loss=0.297, pruned_loss=0.06858, over 8440.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2987, pruned_loss=0.0706, over 1607222.49 frames. ], batch size: 27, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:57:43,371 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 2.421e+02 2.910e+02 3.813e+02 6.624e+02, threshold=5.820e+02, percent-clipped=4.0 +2023-02-06 16:57:58,592 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118186.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:58:17,622 INFO [train.py:901] (1/4) Epoch 15, batch 5050, loss[loss=0.2121, simple_loss=0.3064, pruned_loss=0.05895, over 8510.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2986, pruned_loss=0.07075, over 1606328.76 frames. ], batch size: 28, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:58:23,569 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 16:58:43,446 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 16:58:52,552 INFO [train.py:901] (1/4) Epoch 15, batch 5100, loss[loss=0.2051, simple_loss=0.2892, pruned_loss=0.06044, over 7976.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2996, pruned_loss=0.0711, over 1611982.55 frames. ], batch size: 21, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:58:53,829 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.797e+02 2.591e+02 3.125e+02 3.877e+02 7.785e+02, threshold=6.249e+02, percent-clipped=4.0 +2023-02-06 16:58:57,514 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 16:59:09,116 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118287.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:59:23,820 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118307.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:59:27,781 INFO [train.py:901] (1/4) Epoch 15, batch 5150, loss[loss=0.218, simple_loss=0.2985, pruned_loss=0.06874, over 8292.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2994, pruned_loss=0.07132, over 1607774.76 frames. ], batch size: 23, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:59:51,112 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:00:02,420 INFO [train.py:901] (1/4) Epoch 15, batch 5200, loss[loss=0.1864, simple_loss=0.2554, pruned_loss=0.05869, over 7526.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3, pruned_loss=0.07188, over 1607583.23 frames. ], batch size: 18, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 17:00:03,701 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.390e+02 2.269e+02 2.811e+02 3.673e+02 9.088e+02, threshold=5.623e+02, percent-clipped=2.0 +2023-02-06 17:00:29,669 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118402.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:00:37,916 INFO [train.py:901] (1/4) Epoch 15, batch 5250, loss[loss=0.246, simple_loss=0.3211, pruned_loss=0.08545, over 8611.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2988, pruned_loss=0.07118, over 1607396.34 frames. ], batch size: 31, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 17:00:46,149 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 17:00:56,282 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5515, 2.0139, 3.2511, 1.3641, 2.3310, 1.9018, 1.5825, 2.4133], + device='cuda:1'), covar=tensor([0.1847, 0.2284, 0.0701, 0.4206, 0.1704, 0.2946, 0.2104, 0.2078], + device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0550, 0.0534, 0.0603, 0.0623, 0.0563, 0.0493, 0.0614], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:01:02,215 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3178, 1.7535, 1.2762, 2.6984, 1.3260, 1.0850, 1.9088, 1.9620], + device='cuda:1'), covar=tensor([0.1866, 0.1499, 0.2270, 0.0453, 0.1526, 0.2384, 0.1247, 0.1266], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0203, 0.0248, 0.0211, 0.0211, 0.0245, 0.0253, 0.0212], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 17:01:08,330 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8411, 3.7657, 3.4476, 1.8120, 3.3882, 3.4261, 3.5122, 3.2323], + device='cuda:1'), covar=tensor([0.0886, 0.0601, 0.1047, 0.4499, 0.0889, 0.1170, 0.1209, 0.0955], + device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0400, 0.0407, 0.0503, 0.0399, 0.0409, 0.0387, 0.0352], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:01:12,972 INFO [train.py:901] (1/4) Epoch 15, batch 5300, loss[loss=0.1952, simple_loss=0.2742, pruned_loss=0.0581, over 7664.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3003, pruned_loss=0.07194, over 1609946.31 frames. ], batch size: 19, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 17:01:14,346 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.534e+02 2.995e+02 3.765e+02 8.916e+02, threshold=5.991e+02, percent-clipped=4.0 +2023-02-06 17:01:28,654 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5499, 2.7625, 1.7753, 2.1685, 2.2235, 1.5602, 2.1094, 2.0882], + device='cuda:1'), covar=tensor([0.1568, 0.0360, 0.1159, 0.0753, 0.0763, 0.1463, 0.1023, 0.1081], + device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0228, 0.0323, 0.0298, 0.0299, 0.0327, 0.0342, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:01:47,933 INFO [train.py:901] (1/4) Epoch 15, batch 5350, loss[loss=0.2026, simple_loss=0.2741, pruned_loss=0.06554, over 7543.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2997, pruned_loss=0.07183, over 1608899.16 frames. ], batch size: 18, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 17:01:50,891 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118517.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:02:01,058 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118530.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:02:24,459 INFO [train.py:901] (1/4) Epoch 15, batch 5400, loss[loss=0.2225, simple_loss=0.2927, pruned_loss=0.07615, over 7804.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2993, pruned_loss=0.07151, over 1611614.21 frames. ], batch size: 19, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:02:25,794 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.478e+02 2.903e+02 3.717e+02 8.291e+02, threshold=5.806e+02, percent-clipped=5.0 +2023-02-06 17:02:58,963 INFO [train.py:901] (1/4) Epoch 15, batch 5450, loss[loss=0.2499, simple_loss=0.322, pruned_loss=0.08888, over 8488.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2997, pruned_loss=0.07187, over 1610736.32 frames. ], batch size: 29, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:03:11,227 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:03:21,561 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:03:23,679 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6856, 1.8243, 1.6110, 2.2944, 1.0585, 1.3757, 1.6769, 1.9088], + device='cuda:1'), covar=tensor([0.0751, 0.0714, 0.0948, 0.0421, 0.1017, 0.1344, 0.0755, 0.0656], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0204, 0.0250, 0.0213, 0.0212, 0.0247, 0.0254, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 17:03:24,943 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:03:26,161 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118651.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:03:34,867 INFO [train.py:901] (1/4) Epoch 15, batch 5500, loss[loss=0.2334, simple_loss=0.317, pruned_loss=0.07496, over 8701.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3008, pruned_loss=0.07236, over 1615461.23 frames. ], batch size: 34, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:03:36,245 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.663e+02 2.592e+02 3.113e+02 3.610e+02 8.755e+02, threshold=6.227e+02, percent-clipped=2.0 +2023-02-06 17:03:38,394 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 17:03:54,399 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:04:09,097 INFO [train.py:901] (1/4) Epoch 15, batch 5550, loss[loss=0.1758, simple_loss=0.2562, pruned_loss=0.04768, over 7795.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2994, pruned_loss=0.07155, over 1617728.62 frames. ], batch size: 19, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:04:26,173 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.11 vs. limit=5.0 +2023-02-06 17:04:29,567 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7213, 2.0870, 2.2661, 1.2867, 2.3180, 1.5898, 0.7049, 1.9104], + device='cuda:1'), covar=tensor([0.0506, 0.0236, 0.0239, 0.0526, 0.0335, 0.0690, 0.0715, 0.0264], + device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0357, 0.0307, 0.0412, 0.0342, 0.0501, 0.0370, 0.0379], + device='cuda:1'), out_proj_covar=tensor([1.1430e-04, 9.5911e-05, 8.2284e-05, 1.1148e-04, 9.2555e-05, 1.4585e-04, + 1.0180e-04, 1.0298e-04], device='cuda:1') +2023-02-06 17:04:32,286 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118746.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:04:44,988 INFO [train.py:901] (1/4) Epoch 15, batch 5600, loss[loss=0.2502, simple_loss=0.3265, pruned_loss=0.08693, over 8337.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3002, pruned_loss=0.07127, over 1621938.91 frames. ], batch size: 25, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:04:46,299 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.537e+02 3.218e+02 3.925e+02 9.216e+02, threshold=6.435e+02, percent-clipped=4.0 +2023-02-06 17:04:47,203 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118766.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:04:52,581 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118773.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:04:56,549 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4842, 2.3037, 3.2736, 2.1033, 2.8397, 3.5845, 3.4916, 3.2603], + device='cuda:1'), covar=tensor([0.0754, 0.1099, 0.0483, 0.1506, 0.1026, 0.0187, 0.0567, 0.0373], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0306, 0.0269, 0.0300, 0.0287, 0.0247, 0.0374, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:05:09,079 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118798.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:05:14,494 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118806.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:05:19,155 INFO [train.py:901] (1/4) Epoch 15, batch 5650, loss[loss=0.2571, simple_loss=0.3263, pruned_loss=0.09397, over 8452.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3007, pruned_loss=0.07186, over 1622656.19 frames. ], batch size: 27, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:05:43,439 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 17:05:53,491 INFO [train.py:901] (1/4) Epoch 15, batch 5700, loss[loss=0.1896, simple_loss=0.2605, pruned_loss=0.05939, over 7538.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2995, pruned_loss=0.07132, over 1617099.82 frames. ], batch size: 18, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:05:54,817 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.491e+02 2.972e+02 3.726e+02 7.690e+02, threshold=5.944e+02, percent-clipped=5.0 +2023-02-06 17:05:57,427 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-06 17:06:21,157 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118901.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:06:29,112 INFO [train.py:901] (1/4) Epoch 15, batch 5750, loss[loss=0.2096, simple_loss=0.2826, pruned_loss=0.06832, over 7810.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.299, pruned_loss=0.07113, over 1617104.96 frames. ], batch size: 20, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:06:38,212 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118926.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:06:46,363 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 17:06:51,335 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6601, 1.5664, 2.0674, 1.3951, 1.0927, 2.0774, 0.4057, 1.1867], + device='cuda:1'), covar=tensor([0.2048, 0.1403, 0.0472, 0.1450, 0.3514, 0.0449, 0.2668, 0.1658], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0175, 0.0110, 0.0210, 0.0254, 0.0113, 0.0160, 0.0171], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 17:07:04,191 INFO [train.py:901] (1/4) Epoch 15, batch 5800, loss[loss=0.1751, simple_loss=0.2444, pruned_loss=0.05292, over 7710.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2981, pruned_loss=0.07021, over 1616747.84 frames. ], batch size: 18, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:07:05,538 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.317e+02 2.944e+02 4.100e+02 6.996e+02, threshold=5.887e+02, percent-clipped=4.0 +2023-02-06 17:07:26,177 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:07:32,104 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119002.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:07:39,877 INFO [train.py:901] (1/4) Epoch 15, batch 5850, loss[loss=0.2256, simple_loss=0.3083, pruned_loss=0.07151, over 8355.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2981, pruned_loss=0.06997, over 1619192.21 frames. ], batch size: 24, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:07:46,177 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:07:49,460 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119027.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:02,801 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119047.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:13,677 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:14,146 INFO [train.py:901] (1/4) Epoch 15, batch 5900, loss[loss=0.2207, simple_loss=0.3064, pruned_loss=0.06747, over 8360.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2987, pruned_loss=0.07035, over 1620681.16 frames. ], batch size: 24, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:08:15,366 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.486e+02 2.938e+02 3.942e+02 7.909e+02, threshold=5.877e+02, percent-clipped=6.0 +2023-02-06 17:08:30,156 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:34,144 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:44,788 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119108.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:46,724 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119111.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:47,875 INFO [train.py:901] (1/4) Epoch 15, batch 5950, loss[loss=0.1832, simple_loss=0.2696, pruned_loss=0.04836, over 8246.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2981, pruned_loss=0.07027, over 1622616.55 frames. ], batch size: 22, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:09:21,119 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4840, 1.3665, 2.2865, 1.1938, 1.9127, 2.4377, 2.5558, 2.0706], + device='cuda:1'), covar=tensor([0.0917, 0.1276, 0.0482, 0.2104, 0.0859, 0.0403, 0.0681, 0.0782], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0306, 0.0269, 0.0299, 0.0286, 0.0247, 0.0374, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:09:22,903 INFO [train.py:901] (1/4) Epoch 15, batch 6000, loss[loss=0.2619, simple_loss=0.3298, pruned_loss=0.09697, over 7439.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2978, pruned_loss=0.07019, over 1615457.11 frames. ], batch size: 17, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:09:22,903 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 17:09:35,678 INFO [train.py:935] (1/4) Epoch 15, validation: loss=0.181, simple_loss=0.2808, pruned_loss=0.04056, over 944034.00 frames. +2023-02-06 17:09:35,679 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 17:09:37,104 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.578e+02 3.120e+02 3.956e+02 1.218e+03, threshold=6.240e+02, percent-clipped=5.0 +2023-02-06 17:10:10,481 INFO [train.py:901] (1/4) Epoch 15, batch 6050, loss[loss=0.2912, simple_loss=0.3496, pruned_loss=0.1164, over 8573.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2971, pruned_loss=0.06982, over 1609977.12 frames. ], batch size: 49, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:10:35,628 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.51 vs. limit=5.0 +2023-02-06 17:10:44,316 INFO [train.py:901] (1/4) Epoch 15, batch 6100, loss[loss=0.193, simple_loss=0.2714, pruned_loss=0.05728, over 8071.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2974, pruned_loss=0.07001, over 1610937.32 frames. ], batch size: 21, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:10:45,651 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.463e+02 3.114e+02 4.132e+02 8.492e+02, threshold=6.229e+02, percent-clipped=7.0 +2023-02-06 17:10:58,745 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0216, 2.3099, 1.9852, 2.9137, 1.3837, 1.7514, 2.2295, 2.4526], + device='cuda:1'), covar=tensor([0.0764, 0.0816, 0.0864, 0.0338, 0.1139, 0.1298, 0.0846, 0.0721], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0204, 0.0249, 0.0213, 0.0212, 0.0250, 0.0255, 0.0215], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 17:11:18,272 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 17:11:20,332 INFO [train.py:901] (1/4) Epoch 15, batch 6150, loss[loss=0.2523, simple_loss=0.3397, pruned_loss=0.08243, over 8706.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2977, pruned_loss=0.07009, over 1612006.41 frames. ], batch size: 34, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:11:21,267 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3144, 2.8970, 2.3113, 3.9331, 1.8030, 2.0551, 2.6192, 3.1681], + device='cuda:1'), covar=tensor([0.0771, 0.0811, 0.0892, 0.0238, 0.1115, 0.1299, 0.0917, 0.0756], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0204, 0.0249, 0.0213, 0.0213, 0.0250, 0.0256, 0.0215], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 17:11:41,511 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 17:11:54,697 INFO [train.py:901] (1/4) Epoch 15, batch 6200, loss[loss=0.2681, simple_loss=0.3379, pruned_loss=0.09915, over 8461.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2973, pruned_loss=0.06974, over 1613379.57 frames. ], batch size: 49, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:11:55,639 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:11:56,078 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.347e+02 3.204e+02 3.871e+02 7.576e+02, threshold=6.408e+02, percent-clipped=2.0 +2023-02-06 17:12:10,072 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-02-06 17:12:14,471 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:12:30,389 INFO [train.py:901] (1/4) Epoch 15, batch 6250, loss[loss=0.2105, simple_loss=0.2829, pruned_loss=0.06904, over 7930.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2974, pruned_loss=0.06988, over 1613666.82 frames. ], batch size: 20, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:12:37,235 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2392, 2.2257, 1.6407, 1.9452, 1.7922, 1.4292, 1.6450, 1.6776], + device='cuda:1'), covar=tensor([0.1107, 0.0320, 0.1060, 0.0502, 0.0707, 0.1290, 0.0886, 0.0683], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0231, 0.0326, 0.0302, 0.0302, 0.0331, 0.0346, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:12:47,167 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=119437.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:12:59,472 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=119455.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:13:04,854 INFO [train.py:901] (1/4) Epoch 15, batch 6300, loss[loss=0.1771, simple_loss=0.2576, pruned_loss=0.04826, over 7804.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2976, pruned_loss=0.07011, over 1614490.15 frames. ], batch size: 20, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:13:06,146 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.517e+02 3.087e+02 3.932e+02 1.134e+03, threshold=6.173e+02, percent-clipped=3.0 +2023-02-06 17:13:41,045 INFO [train.py:901] (1/4) Epoch 15, batch 6350, loss[loss=0.2574, simple_loss=0.3445, pruned_loss=0.08514, over 8252.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2985, pruned_loss=0.07069, over 1612442.25 frames. ], batch size: 24, lr: 5.03e-03, grad_scale: 8.0 +2023-02-06 17:13:53,782 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:13:57,576 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.41 vs. limit=5.0 +2023-02-06 17:14:07,838 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119552.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:14:15,091 INFO [train.py:901] (1/4) Epoch 15, batch 6400, loss[loss=0.2024, simple_loss=0.2775, pruned_loss=0.06363, over 8356.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2988, pruned_loss=0.07128, over 1610513.78 frames. ], batch size: 24, lr: 5.03e-03, grad_scale: 8.0 +2023-02-06 17:14:16,446 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.423e+02 3.023e+02 3.752e+02 7.818e+02, threshold=6.047e+02, percent-clipped=4.0 +2023-02-06 17:14:20,028 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119570.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:14:45,443 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6663, 5.8391, 5.0614, 2.3194, 5.0655, 5.4028, 5.3545, 5.0764], + device='cuda:1'), covar=tensor([0.0532, 0.0365, 0.0931, 0.4353, 0.0734, 0.0826, 0.1003, 0.0593], + device='cuda:1'), in_proj_covar=tensor([0.0489, 0.0405, 0.0408, 0.0502, 0.0398, 0.0411, 0.0389, 0.0354], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:14:49,984 INFO [train.py:901] (1/4) Epoch 15, batch 6450, loss[loss=0.2101, simple_loss=0.299, pruned_loss=0.06058, over 8458.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2987, pruned_loss=0.07103, over 1611966.43 frames. ], batch size: 27, lr: 5.03e-03, grad_scale: 8.0 +2023-02-06 17:15:24,247 INFO [train.py:901] (1/4) Epoch 15, batch 6500, loss[loss=0.2111, simple_loss=0.2728, pruned_loss=0.07465, over 7548.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.299, pruned_loss=0.0709, over 1615813.01 frames. ], batch size: 18, lr: 5.03e-03, grad_scale: 8.0 +2023-02-06 17:15:25,571 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.561e+02 2.888e+02 3.578e+02 6.995e+02, threshold=5.776e+02, percent-clipped=4.0 +2023-02-06 17:15:38,583 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119683.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:15:55,014 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 17:15:58,717 INFO [train.py:901] (1/4) Epoch 15, batch 6550, loss[loss=0.2163, simple_loss=0.2995, pruned_loss=0.06657, over 7675.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2984, pruned_loss=0.0703, over 1615739.82 frames. ], batch size: 19, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:16:29,740 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 17:16:34,450 INFO [train.py:901] (1/4) Epoch 15, batch 6600, loss[loss=0.2328, simple_loss=0.3119, pruned_loss=0.07686, over 8446.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.299, pruned_loss=0.06976, over 1620644.73 frames. ], batch size: 27, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:16:35,799 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 2.456e+02 2.938e+02 3.854e+02 9.901e+02, threshold=5.877e+02, percent-clipped=5.0 +2023-02-06 17:16:47,919 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 17:17:05,449 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119808.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:17:08,629 INFO [train.py:901] (1/4) Epoch 15, batch 6650, loss[loss=0.2049, simple_loss=0.2853, pruned_loss=0.0622, over 7912.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2988, pruned_loss=0.06976, over 1617160.45 frames. ], batch size: 20, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:17:15,023 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0479, 1.2712, 1.1986, 0.7032, 1.2157, 1.0422, 0.1002, 1.2082], + device='cuda:1'), covar=tensor([0.0328, 0.0281, 0.0239, 0.0404, 0.0310, 0.0738, 0.0639, 0.0250], + device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0364, 0.0314, 0.0421, 0.0348, 0.0506, 0.0376, 0.0387], + device='cuda:1'), out_proj_covar=tensor([1.1614e-04, 9.7785e-05, 8.4111e-05, 1.1365e-04, 9.4181e-05, 1.4720e-04, + 1.0351e-04, 1.0515e-04], device='cuda:1') +2023-02-06 17:17:17,671 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119826.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:17:22,315 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:17:36,355 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:17:44,400 INFO [train.py:901] (1/4) Epoch 15, batch 6700, loss[loss=0.1665, simple_loss=0.2551, pruned_loss=0.03899, over 7548.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2994, pruned_loss=0.0699, over 1622549.46 frames. ], batch size: 18, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:17:45,218 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7785, 5.8618, 5.1026, 2.4854, 5.2106, 5.6258, 5.4364, 5.3645], + device='cuda:1'), covar=tensor([0.0522, 0.0359, 0.0837, 0.4109, 0.0657, 0.0860, 0.0970, 0.0687], + device='cuda:1'), in_proj_covar=tensor([0.0489, 0.0405, 0.0412, 0.0504, 0.0402, 0.0409, 0.0392, 0.0355], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:17:45,753 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.601e+02 2.951e+02 3.516e+02 8.618e+02, threshold=5.902e+02, percent-clipped=2.0 +2023-02-06 17:17:52,175 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8872, 1.6794, 3.2724, 1.5395, 2.3907, 3.6029, 3.6973, 3.0204], + device='cuda:1'), covar=tensor([0.1192, 0.1585, 0.0311, 0.2008, 0.0914, 0.0233, 0.0535, 0.0587], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0309, 0.0272, 0.0302, 0.0288, 0.0250, 0.0380, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:17:53,450 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=119876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:18:19,548 INFO [train.py:901] (1/4) Epoch 15, batch 6750, loss[loss=0.2331, simple_loss=0.3085, pruned_loss=0.07884, over 8300.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2987, pruned_loss=0.0697, over 1620373.92 frames. ], batch size: 23, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:18:19,807 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5182, 1.9602, 2.9953, 1.3189, 2.2650, 1.8933, 1.5992, 2.1903], + device='cuda:1'), covar=tensor([0.1916, 0.2288, 0.0837, 0.4251, 0.1622, 0.3052, 0.2114, 0.2141], + device='cuda:1'), in_proj_covar=tensor([0.0502, 0.0558, 0.0539, 0.0607, 0.0629, 0.0569, 0.0499, 0.0621], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:18:55,246 INFO [train.py:901] (1/4) Epoch 15, batch 6800, loss[loss=0.2048, simple_loss=0.2855, pruned_loss=0.06208, over 8241.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2986, pruned_loss=0.06934, over 1620244.13 frames. ], batch size: 22, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:18:57,360 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.559e+02 3.032e+02 3.835e+02 7.300e+02, threshold=6.064e+02, percent-clipped=2.0 +2023-02-06 17:19:03,570 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 17:19:15,432 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119991.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:19:32,093 INFO [train.py:901] (1/4) Epoch 15, batch 6850, loss[loss=0.2314, simple_loss=0.3227, pruned_loss=0.07003, over 8035.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.298, pruned_loss=0.06876, over 1619004.57 frames. ], batch size: 22, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:19:41,652 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=120027.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:19:53,396 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 17:20:06,203 INFO [train.py:901] (1/4) Epoch 15, batch 6900, loss[loss=0.2005, simple_loss=0.2785, pruned_loss=0.06127, over 8092.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2967, pruned_loss=0.06808, over 1617990.11 frames. ], batch size: 21, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:20:07,530 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.397e+02 2.973e+02 3.506e+02 9.980e+02, threshold=5.947e+02, percent-clipped=2.0 +2023-02-06 17:20:42,260 INFO [train.py:901] (1/4) Epoch 15, batch 6950, loss[loss=0.2356, simple_loss=0.3105, pruned_loss=0.08033, over 8247.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2972, pruned_loss=0.06854, over 1617800.78 frames. ], batch size: 24, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:21:02,382 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120142.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:21:03,577 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 17:21:16,268 INFO [train.py:901] (1/4) Epoch 15, batch 7000, loss[loss=0.2048, simple_loss=0.2827, pruned_loss=0.06344, over 8237.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2977, pruned_loss=0.06926, over 1616812.78 frames. ], batch size: 22, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:21:17,612 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.303e+02 2.879e+02 3.620e+02 6.461e+02, threshold=5.757e+02, percent-clipped=3.0 +2023-02-06 17:21:51,892 INFO [train.py:901] (1/4) Epoch 15, batch 7050, loss[loss=0.2502, simple_loss=0.3401, pruned_loss=0.0801, over 8582.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2984, pruned_loss=0.06995, over 1614972.78 frames. ], batch size: 31, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:22:11,821 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-06 17:22:15,019 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:22:26,161 INFO [train.py:901] (1/4) Epoch 15, batch 7100, loss[loss=0.2217, simple_loss=0.3006, pruned_loss=0.07139, over 8189.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2978, pruned_loss=0.06972, over 1619074.86 frames. ], batch size: 23, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:22:27,488 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.428e+02 3.078e+02 4.147e+02 9.225e+02, threshold=6.156e+02, percent-clipped=10.0 +2023-02-06 17:22:32,365 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120272.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:22:57,827 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.15 vs. limit=5.0 +2023-02-06 17:23:00,855 INFO [train.py:901] (1/4) Epoch 15, batch 7150, loss[loss=0.2268, simple_loss=0.3129, pruned_loss=0.07034, over 8475.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2988, pruned_loss=0.07002, over 1621836.96 frames. ], batch size: 25, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:23:27,589 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6404, 1.3732, 1.6093, 1.2252, 0.8730, 1.4749, 1.4750, 1.2335], + device='cuda:1'), covar=tensor([0.0509, 0.1232, 0.1623, 0.1447, 0.0600, 0.1462, 0.0685, 0.0657], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0156, 0.0100, 0.0162, 0.0114, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 17:23:27,658 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5115, 1.8622, 2.7434, 1.3786, 2.0274, 1.8798, 1.6218, 1.9170], + device='cuda:1'), covar=tensor([0.1734, 0.2193, 0.0798, 0.4196, 0.1540, 0.2836, 0.1891, 0.2006], + device='cuda:1'), in_proj_covar=tensor([0.0499, 0.0552, 0.0537, 0.0603, 0.0626, 0.0565, 0.0495, 0.0619], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:23:31,187 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 17:23:35,462 INFO [train.py:901] (1/4) Epoch 15, batch 7200, loss[loss=0.237, simple_loss=0.3194, pruned_loss=0.07733, over 8335.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2988, pruned_loss=0.07011, over 1621641.50 frames. ], batch size: 25, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:23:36,809 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.418e+02 2.853e+02 3.692e+02 6.645e+02, threshold=5.707e+02, percent-clipped=2.0 +2023-02-06 17:24:00,217 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:24:10,187 INFO [train.py:901] (1/4) Epoch 15, batch 7250, loss[loss=0.2278, simple_loss=0.3108, pruned_loss=0.07239, over 8581.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2984, pruned_loss=0.07005, over 1620435.41 frames. ], batch size: 39, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:24:17,873 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120423.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:24:29,239 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4643, 1.8387, 2.7164, 1.2994, 1.8113, 1.7786, 1.5633, 1.8661], + device='cuda:1'), covar=tensor([0.1808, 0.2271, 0.0788, 0.4291, 0.1834, 0.3059, 0.2086, 0.2173], + device='cuda:1'), in_proj_covar=tensor([0.0501, 0.0554, 0.0539, 0.0609, 0.0629, 0.0568, 0.0498, 0.0621], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:24:45,978 INFO [train.py:901] (1/4) Epoch 15, batch 7300, loss[loss=0.2139, simple_loss=0.3003, pruned_loss=0.06372, over 8510.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2989, pruned_loss=0.07007, over 1620851.55 frames. ], batch size: 28, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:24:47,334 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.423e+02 2.925e+02 3.483e+02 5.889e+02, threshold=5.849e+02, percent-clipped=3.0 +2023-02-06 17:25:18,671 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3409, 1.5550, 4.4158, 1.8565, 2.2344, 5.1214, 5.0906, 4.3455], + device='cuda:1'), covar=tensor([0.1070, 0.1799, 0.0253, 0.1855, 0.1208, 0.0155, 0.0347, 0.0544], + device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0306, 0.0272, 0.0301, 0.0285, 0.0248, 0.0379, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:25:20,528 INFO [train.py:901] (1/4) Epoch 15, batch 7350, loss[loss=0.1792, simple_loss=0.2626, pruned_loss=0.04784, over 7803.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2989, pruned_loss=0.07008, over 1621910.99 frames. ], batch size: 19, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:25:27,942 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0087, 1.7065, 3.4272, 1.4492, 2.2906, 3.8793, 3.8876, 3.2817], + device='cuda:1'), covar=tensor([0.1113, 0.1582, 0.0349, 0.2030, 0.1018, 0.0208, 0.0490, 0.0576], + device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0306, 0.0271, 0.0301, 0.0285, 0.0248, 0.0379, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:25:45,373 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 17:25:56,247 INFO [train.py:901] (1/4) Epoch 15, batch 7400, loss[loss=0.22, simple_loss=0.2933, pruned_loss=0.07332, over 7982.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2999, pruned_loss=0.07101, over 1618677.54 frames. ], batch size: 21, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:25:57,546 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.487e+02 3.190e+02 4.160e+02 9.613e+02, threshold=6.380e+02, percent-clipped=9.0 +2023-02-06 17:26:04,623 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 17:26:18,149 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2568, 1.4299, 1.5331, 1.2108, 0.8933, 1.4068, 1.7928, 1.7456], + device='cuda:1'), covar=tensor([0.0445, 0.1295, 0.1771, 0.1501, 0.0592, 0.1573, 0.0637, 0.0600], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0156, 0.0099, 0.0161, 0.0113, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 17:26:30,893 INFO [train.py:901] (1/4) Epoch 15, batch 7450, loss[loss=0.2022, simple_loss=0.2944, pruned_loss=0.05496, over 8017.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2992, pruned_loss=0.07084, over 1616912.25 frames. ], batch size: 22, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:26:42,795 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 17:27:06,396 INFO [train.py:901] (1/4) Epoch 15, batch 7500, loss[loss=0.1912, simple_loss=0.2818, pruned_loss=0.05028, over 8635.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2984, pruned_loss=0.0707, over 1612834.43 frames. ], batch size: 31, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:27:07,754 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.388e+02 2.853e+02 3.831e+02 7.536e+02, threshold=5.707e+02, percent-clipped=4.0 +2023-02-06 17:27:27,369 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120694.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:27:37,317 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-06 17:27:40,251 INFO [train.py:901] (1/4) Epoch 15, batch 7550, loss[loss=0.2178, simple_loss=0.295, pruned_loss=0.07029, over 8228.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2995, pruned_loss=0.07122, over 1619823.19 frames. ], batch size: 22, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:28:14,835 INFO [train.py:901] (1/4) Epoch 15, batch 7600, loss[loss=0.1912, simple_loss=0.2658, pruned_loss=0.05834, over 7437.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2975, pruned_loss=0.0703, over 1613283.14 frames. ], batch size: 17, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:28:16,204 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.426e+02 3.048e+02 3.965e+02 8.844e+02, threshold=6.096e+02, percent-clipped=6.0 +2023-02-06 17:28:50,148 INFO [train.py:901] (1/4) Epoch 15, batch 7650, loss[loss=0.1965, simple_loss=0.2896, pruned_loss=0.05165, over 8187.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.299, pruned_loss=0.07111, over 1615516.01 frames. ], batch size: 23, lr: 5.01e-03, grad_scale: 8.0 +2023-02-06 17:29:25,339 INFO [train.py:901] (1/4) Epoch 15, batch 7700, loss[loss=0.2361, simple_loss=0.313, pruned_loss=0.07964, over 8515.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2981, pruned_loss=0.07074, over 1612962.91 frames. ], batch size: 28, lr: 5.01e-03, grad_scale: 8.0 +2023-02-06 17:29:27,394 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.773e+02 2.503e+02 3.087e+02 4.175e+02 9.539e+02, threshold=6.174e+02, percent-clipped=7.0 +2023-02-06 17:29:40,156 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 17:29:52,764 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 17:30:01,558 INFO [train.py:901] (1/4) Epoch 15, batch 7750, loss[loss=0.1723, simple_loss=0.2495, pruned_loss=0.04756, over 7541.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2962, pruned_loss=0.0691, over 1612152.45 frames. ], batch size: 18, lr: 5.01e-03, grad_scale: 8.0 +2023-02-06 17:30:36,087 INFO [train.py:901] (1/4) Epoch 15, batch 7800, loss[loss=0.2053, simple_loss=0.2962, pruned_loss=0.05715, over 8657.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2965, pruned_loss=0.06912, over 1613952.68 frames. ], batch size: 39, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:30:38,109 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.376e+02 2.783e+02 3.266e+02 5.993e+02, threshold=5.565e+02, percent-clipped=0.0 +2023-02-06 17:30:55,826 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5978, 1.8723, 3.0054, 1.3512, 2.2495, 1.8906, 1.6792, 2.0723], + device='cuda:1'), covar=tensor([0.1568, 0.2042, 0.0670, 0.3707, 0.1389, 0.2735, 0.1746, 0.2103], + device='cuda:1'), in_proj_covar=tensor([0.0498, 0.0551, 0.0536, 0.0604, 0.0624, 0.0567, 0.0497, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:31:09,465 INFO [train.py:901] (1/4) Epoch 15, batch 7850, loss[loss=0.2059, simple_loss=0.2938, pruned_loss=0.05904, over 8468.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.299, pruned_loss=0.07077, over 1617713.68 frames. ], batch size: 25, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:31:14,866 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121021.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:31:26,042 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=121038.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:31:42,572 INFO [train.py:901] (1/4) Epoch 15, batch 7900, loss[loss=0.1929, simple_loss=0.272, pruned_loss=0.05697, over 7666.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3002, pruned_loss=0.07168, over 1619384.06 frames. ], batch size: 19, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:31:44,513 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.419e+02 3.139e+02 4.114e+02 1.036e+03, threshold=6.279e+02, percent-clipped=8.0 +2023-02-06 17:31:45,339 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9778, 1.8538, 6.0651, 2.3275, 5.5406, 5.1706, 5.6264, 5.5925], + device='cuda:1'), covar=tensor([0.0462, 0.4372, 0.0312, 0.3220, 0.0855, 0.0779, 0.0473, 0.0437], + device='cuda:1'), in_proj_covar=tensor([0.0558, 0.0610, 0.0630, 0.0579, 0.0652, 0.0564, 0.0555, 0.0613], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 17:32:05,284 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-02-06 17:32:15,966 INFO [train.py:901] (1/4) Epoch 15, batch 7950, loss[loss=0.2104, simple_loss=0.2831, pruned_loss=0.06882, over 8087.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2982, pruned_loss=0.07033, over 1617570.77 frames. ], batch size: 21, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:32:40,729 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4282, 2.3860, 1.6081, 2.0013, 1.9717, 1.4786, 1.8047, 1.8571], + device='cuda:1'), covar=tensor([0.1389, 0.0367, 0.1263, 0.0583, 0.0655, 0.1401, 0.1009, 0.0866], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0232, 0.0322, 0.0302, 0.0300, 0.0328, 0.0339, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:32:42,026 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:32:48,346 INFO [train.py:901] (1/4) Epoch 15, batch 8000, loss[loss=0.2198, simple_loss=0.2935, pruned_loss=0.07302, over 7646.00 frames. ], tot_loss[loss=0.219, simple_loss=0.298, pruned_loss=0.07004, over 1613903.04 frames. ], batch size: 19, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:32:50,378 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.460e+02 2.992e+02 3.696e+02 7.694e+02, threshold=5.984e+02, percent-clipped=2.0 +2023-02-06 17:33:01,512 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121182.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:33:22,852 INFO [train.py:901] (1/4) Epoch 15, batch 8050, loss[loss=0.1907, simple_loss=0.2691, pruned_loss=0.05614, over 8074.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2963, pruned_loss=0.0702, over 1599456.27 frames. ], batch size: 21, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:33:34,505 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1105, 1.3981, 1.7446, 1.2512, 1.0748, 1.4244, 1.7792, 1.3979], + device='cuda:1'), covar=tensor([0.0457, 0.1228, 0.1606, 0.1377, 0.0564, 0.1468, 0.0645, 0.0708], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0191, 0.0157, 0.0100, 0.0162, 0.0113, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 17:33:55,806 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 17:34:00,933 INFO [train.py:901] (1/4) Epoch 16, batch 0, loss[loss=0.2322, simple_loss=0.3085, pruned_loss=0.0779, over 7283.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3085, pruned_loss=0.0779, over 7283.00 frames. ], batch size: 71, lr: 4.84e-03, grad_scale: 8.0 +2023-02-06 17:34:00,934 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 17:34:11,350 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5955, 1.5610, 2.6884, 1.2832, 1.9959, 2.8710, 3.0526, 2.4475], + device='cuda:1'), covar=tensor([0.1262, 0.1554, 0.0450, 0.2407, 0.0900, 0.0403, 0.0826, 0.0812], + device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0309, 0.0271, 0.0301, 0.0288, 0.0248, 0.0377, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:34:11,915 INFO [train.py:935] (1/4) Epoch 16, validation: loss=0.1795, simple_loss=0.2801, pruned_loss=0.03944, over 944034.00 frames. +2023-02-06 17:34:11,916 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 17:34:17,798 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4994, 1.4235, 4.6868, 1.7023, 4.0936, 3.8633, 4.1505, 4.0700], + device='cuda:1'), covar=tensor([0.0504, 0.4903, 0.0452, 0.4032, 0.1121, 0.1044, 0.0610, 0.0666], + device='cuda:1'), in_proj_covar=tensor([0.0558, 0.0611, 0.0632, 0.0582, 0.0653, 0.0565, 0.0555, 0.0614], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 17:34:24,910 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.543e+02 3.194e+02 4.084e+02 8.334e+02, threshold=6.389e+02, percent-clipped=7.0 +2023-02-06 17:34:26,234 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 17:34:41,509 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 +2023-02-06 17:34:47,407 INFO [train.py:901] (1/4) Epoch 16, batch 50, loss[loss=0.2155, simple_loss=0.2961, pruned_loss=0.06745, over 8246.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2958, pruned_loss=0.06951, over 361301.50 frames. ], batch size: 22, lr: 4.84e-03, grad_scale: 8.0 +2023-02-06 17:35:00,270 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 17:35:02,266 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 17:35:09,789 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121329.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 17:35:20,724 INFO [train.py:901] (1/4) Epoch 16, batch 100, loss[loss=0.2028, simple_loss=0.2769, pruned_loss=0.06435, over 8034.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3002, pruned_loss=0.07091, over 637782.19 frames. ], batch size: 22, lr: 4.84e-03, grad_scale: 8.0 +2023-02-06 17:35:24,733 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 17:35:33,278 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=121365.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:35:33,867 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.470e+02 2.913e+02 3.674e+02 6.203e+02, threshold=5.826e+02, percent-clipped=0.0 +2023-02-06 17:35:52,048 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5755, 2.1076, 3.4385, 1.3985, 2.5428, 2.0290, 1.6714, 2.5940], + device='cuda:1'), covar=tensor([0.1791, 0.2403, 0.0715, 0.4115, 0.1671, 0.3094, 0.2101, 0.2085], + device='cuda:1'), in_proj_covar=tensor([0.0508, 0.0560, 0.0542, 0.0614, 0.0637, 0.0580, 0.0507, 0.0627], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:35:53,856 INFO [train.py:901] (1/4) Epoch 16, batch 150, loss[loss=0.2541, simple_loss=0.3423, pruned_loss=0.08293, over 8493.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2995, pruned_loss=0.06991, over 857519.53 frames. ], batch size: 28, lr: 4.84e-03, grad_scale: 8.0 +2023-02-06 17:36:00,976 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8193, 1.6480, 2.3865, 1.6677, 1.2440, 2.3931, 0.3064, 1.4171], + device='cuda:1'), covar=tensor([0.2246, 0.1743, 0.0521, 0.1725, 0.3711, 0.0496, 0.3041, 0.1784], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0180, 0.0112, 0.0215, 0.0257, 0.0116, 0.0163, 0.0176], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 17:36:04,319 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121409.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:36:15,650 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121425.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:36:21,672 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:36:30,080 INFO [train.py:901] (1/4) Epoch 16, batch 200, loss[loss=0.195, simple_loss=0.2827, pruned_loss=0.05367, over 8283.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3001, pruned_loss=0.07019, over 1028455.03 frames. ], batch size: 23, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:36:33,180 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.88 vs. limit=5.0 +2023-02-06 17:36:43,676 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.385e+02 2.940e+02 3.661e+02 7.455e+02, threshold=5.881e+02, percent-clipped=4.0 +2023-02-06 17:36:53,373 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121480.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:36:56,049 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4986, 2.7970, 2.2490, 3.6912, 1.6762, 1.8448, 2.2090, 3.0059], + device='cuda:1'), covar=tensor([0.0623, 0.0776, 0.0847, 0.0318, 0.1178, 0.1291, 0.1078, 0.0693], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0206, 0.0252, 0.0215, 0.0216, 0.0252, 0.0257, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 17:36:59,471 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3927, 2.4951, 1.7449, 2.0741, 2.0643, 1.5323, 1.9093, 1.8123], + device='cuda:1'), covar=tensor([0.1387, 0.0339, 0.1237, 0.0561, 0.0608, 0.1375, 0.0912, 0.0939], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0233, 0.0326, 0.0303, 0.0301, 0.0330, 0.0341, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:37:04,015 INFO [train.py:901] (1/4) Epoch 16, batch 250, loss[loss=0.2294, simple_loss=0.3084, pruned_loss=0.07524, over 7803.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2992, pruned_loss=0.07039, over 1157534.87 frames. ], batch size: 20, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:37:07,580 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6406, 2.3097, 4.3834, 1.5314, 3.0181, 2.3676, 1.7479, 2.7754], + device='cuda:1'), covar=tensor([0.1900, 0.2493, 0.0691, 0.4344, 0.1707, 0.2976, 0.2135, 0.2477], + device='cuda:1'), in_proj_covar=tensor([0.0507, 0.0559, 0.0542, 0.0612, 0.0636, 0.0578, 0.0505, 0.0627], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:37:18,651 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 17:37:24,808 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=121526.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:37:28,152 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 17:37:39,707 INFO [train.py:901] (1/4) Epoch 16, batch 300, loss[loss=0.2173, simple_loss=0.2916, pruned_loss=0.0715, over 7813.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3, pruned_loss=0.0706, over 1261683.93 frames. ], batch size: 20, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:37:54,072 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.529e+02 3.079e+02 3.820e+02 7.739e+02, threshold=6.158e+02, percent-clipped=5.0 +2023-02-06 17:38:14,118 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8166, 1.8955, 1.6454, 2.2469, 1.0019, 1.4192, 1.6460, 1.9060], + device='cuda:1'), covar=tensor([0.0756, 0.0747, 0.0994, 0.0461, 0.1254, 0.1474, 0.0870, 0.0760], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0205, 0.0250, 0.0214, 0.0214, 0.0251, 0.0254, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 17:38:14,590 INFO [train.py:901] (1/4) Epoch 16, batch 350, loss[loss=0.2279, simple_loss=0.3097, pruned_loss=0.07303, over 8261.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2988, pruned_loss=0.06971, over 1341250.27 frames. ], batch size: 24, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:38:45,978 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121641.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:38:49,789 INFO [train.py:901] (1/4) Epoch 16, batch 400, loss[loss=0.2403, simple_loss=0.3241, pruned_loss=0.07821, over 8450.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2988, pruned_loss=0.07017, over 1397511.31 frames. ], batch size: 29, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:39:04,292 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.467e+02 3.087e+02 3.761e+02 6.357e+02, threshold=6.175e+02, percent-clipped=1.0 +2023-02-06 17:39:09,180 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=121673.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 17:39:25,132 INFO [train.py:901] (1/4) Epoch 16, batch 450, loss[loss=0.2471, simple_loss=0.3227, pruned_loss=0.08572, over 8236.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2978, pruned_loss=0.06943, over 1445370.78 frames. ], batch size: 22, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:39:52,406 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:39:59,044 INFO [train.py:901] (1/4) Epoch 16, batch 500, loss[loss=0.2348, simple_loss=0.2954, pruned_loss=0.08716, over 7257.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2986, pruned_loss=0.06958, over 1483758.34 frames. ], batch size: 16, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:40:00,105 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.00 vs. limit=5.0 +2023-02-06 17:40:10,963 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:40:14,769 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.435e+02 2.838e+02 3.555e+02 6.989e+02, threshold=5.677e+02, percent-clipped=1.0 +2023-02-06 17:40:17,013 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=121769.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:40:29,902 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121788.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 17:40:35,812 INFO [train.py:901] (1/4) Epoch 16, batch 550, loss[loss=0.2323, simple_loss=0.3107, pruned_loss=0.0769, over 8323.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2985, pruned_loss=0.06962, over 1512187.58 frames. ], batch size: 25, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:41:02,682 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9201, 1.3509, 3.2750, 1.3718, 2.2937, 3.5766, 3.6990, 3.0439], + device='cuda:1'), covar=tensor([0.1138, 0.1879, 0.0355, 0.2203, 0.1050, 0.0238, 0.0579, 0.0652], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0310, 0.0273, 0.0301, 0.0292, 0.0248, 0.0381, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:41:09,168 INFO [train.py:901] (1/4) Epoch 16, batch 600, loss[loss=0.2012, simple_loss=0.2661, pruned_loss=0.06815, over 7240.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2976, pruned_loss=0.06899, over 1536578.89 frames. ], batch size: 16, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:41:15,570 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 17:41:22,430 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.425e+02 3.086e+02 4.175e+02 1.417e+03, threshold=6.173e+02, percent-clipped=9.0 +2023-02-06 17:41:26,593 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 17:41:36,804 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121884.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:41:44,746 INFO [train.py:901] (1/4) Epoch 16, batch 650, loss[loss=0.2318, simple_loss=0.3087, pruned_loss=0.0774, over 8505.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.297, pruned_loss=0.06839, over 1552590.39 frames. ], batch size: 28, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:41:45,635 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121897.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:41:48,936 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 17:42:02,836 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121922.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:42:18,673 INFO [train.py:901] (1/4) Epoch 16, batch 700, loss[loss=0.2186, simple_loss=0.2934, pruned_loss=0.07187, over 8245.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2974, pruned_loss=0.06873, over 1568973.22 frames. ], batch size: 22, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:42:32,102 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.351e+02 2.496e+02 2.978e+02 3.542e+02 1.118e+03, threshold=5.957e+02, percent-clipped=1.0 +2023-02-06 17:42:33,596 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121968.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:42:41,116 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 17:42:53,714 INFO [train.py:901] (1/4) Epoch 16, batch 750, loss[loss=0.1422, simple_loss=0.2207, pruned_loss=0.03182, over 7445.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2963, pruned_loss=0.06816, over 1574937.52 frames. ], batch size: 17, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:43:14,269 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 17:43:14,507 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0496, 1.2536, 1.1631, 0.5839, 1.2408, 1.0349, 0.0900, 1.1723], + device='cuda:1'), covar=tensor([0.0352, 0.0314, 0.0265, 0.0523, 0.0350, 0.0811, 0.0671, 0.0290], + device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0355, 0.0306, 0.0409, 0.0341, 0.0495, 0.0361, 0.0377], + device='cuda:1'), out_proj_covar=tensor([1.1340e-04, 9.4978e-05, 8.1782e-05, 1.1026e-04, 9.1961e-05, 1.4363e-04, + 9.9360e-05, 1.0221e-04], device='cuda:1') +2023-02-06 17:43:23,870 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 17:43:28,672 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122044.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 17:43:29,777 INFO [train.py:901] (1/4) Epoch 16, batch 800, loss[loss=0.1845, simple_loss=0.2504, pruned_loss=0.05933, over 7441.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2968, pruned_loss=0.06897, over 1586342.86 frames. ], batch size: 17, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:43:43,081 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.422e+02 2.925e+02 3.576e+02 6.712e+02, threshold=5.851e+02, percent-clipped=2.0 +2023-02-06 17:43:45,459 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122069.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 17:44:03,130 INFO [train.py:901] (1/4) Epoch 16, batch 850, loss[loss=0.1906, simple_loss=0.2706, pruned_loss=0.05532, over 7808.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2974, pruned_loss=0.06902, over 1595202.53 frames. ], batch size: 20, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:44:16,338 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9393, 2.3921, 3.5291, 2.1545, 1.8044, 3.4943, 0.8776, 2.0844], + device='cuda:1'), covar=tensor([0.1603, 0.1447, 0.0309, 0.1787, 0.2947, 0.0333, 0.2666, 0.1832], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0179, 0.0112, 0.0212, 0.0255, 0.0115, 0.0161, 0.0176], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 17:44:31,616 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:44:34,938 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:44:38,752 INFO [train.py:901] (1/4) Epoch 16, batch 900, loss[loss=0.244, simple_loss=0.3202, pruned_loss=0.08389, over 8129.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2971, pruned_loss=0.06881, over 1598158.05 frames. ], batch size: 22, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:44:48,951 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6617, 2.2952, 3.3346, 1.9341, 1.6000, 3.2542, 0.8072, 2.0377], + device='cuda:1'), covar=tensor([0.1754, 0.1408, 0.0292, 0.1953, 0.3185, 0.0345, 0.2743, 0.1521], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0179, 0.0112, 0.0212, 0.0255, 0.0115, 0.0161, 0.0176], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 17:44:52,336 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:44:52,793 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.482e+02 3.085e+02 4.013e+02 7.148e+02, threshold=6.170e+02, percent-clipped=4.0 +2023-02-06 17:45:01,231 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 17:45:12,889 INFO [train.py:901] (1/4) Epoch 16, batch 950, loss[loss=0.2391, simple_loss=0.3155, pruned_loss=0.0814, over 8681.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2968, pruned_loss=0.06844, over 1604579.06 frames. ], batch size: 39, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:45:24,854 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:45:40,117 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 17:45:49,040 INFO [train.py:901] (1/4) Epoch 16, batch 1000, loss[loss=0.1906, simple_loss=0.2741, pruned_loss=0.05352, over 8341.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2968, pruned_loss=0.06818, over 1611459.99 frames. ], batch size: 25, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:46:03,407 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.462e+02 3.004e+02 3.600e+02 8.525e+02, threshold=6.009e+02, percent-clipped=4.0 +2023-02-06 17:46:14,159 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 17:46:23,684 INFO [train.py:901] (1/4) Epoch 16, batch 1050, loss[loss=0.2166, simple_loss=0.3052, pruned_loss=0.06394, over 8348.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2966, pruned_loss=0.06777, over 1616777.83 frames. ], batch size: 26, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:46:26,430 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 17:46:34,620 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:46:51,299 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122337.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:46:57,875 INFO [train.py:901] (1/4) Epoch 16, batch 1100, loss[loss=0.1787, simple_loss=0.2636, pruned_loss=0.04692, over 7932.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2963, pruned_loss=0.06752, over 1616989.05 frames. ], batch size: 20, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:47:12,623 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.699e+02 3.204e+02 3.982e+02 8.590e+02, threshold=6.408e+02, percent-clipped=5.0 +2023-02-06 17:47:33,544 INFO [train.py:901] (1/4) Epoch 16, batch 1150, loss[loss=0.2085, simple_loss=0.2815, pruned_loss=0.06775, over 7222.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.295, pruned_loss=0.06704, over 1615512.79 frames. ], batch size: 16, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:47:38,300 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 17:47:54,848 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122427.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:48:07,544 INFO [train.py:901] (1/4) Epoch 16, batch 1200, loss[loss=0.218, simple_loss=0.2991, pruned_loss=0.06841, over 8587.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2958, pruned_loss=0.06836, over 1614984.24 frames. ], batch size: 31, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:48:21,992 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.417e+02 3.007e+02 3.779e+02 1.089e+03, threshold=6.013e+02, percent-clipped=2.0 +2023-02-06 17:48:31,788 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:48:36,690 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122486.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:48:43,455 INFO [train.py:901] (1/4) Epoch 16, batch 1250, loss[loss=0.2418, simple_loss=0.3097, pruned_loss=0.0869, over 7915.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2956, pruned_loss=0.0685, over 1618450.77 frames. ], batch size: 20, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:49:19,086 INFO [train.py:901] (1/4) Epoch 16, batch 1300, loss[loss=0.1963, simple_loss=0.2817, pruned_loss=0.05546, over 8248.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2957, pruned_loss=0.06883, over 1617056.59 frames. ], batch size: 24, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:49:26,973 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122557.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:49:33,313 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.571e+02 3.105e+02 3.703e+02 6.719e+02, threshold=6.210e+02, percent-clipped=4.0 +2023-02-06 17:49:54,912 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:49:56,060 INFO [train.py:901] (1/4) Epoch 16, batch 1350, loss[loss=0.2459, simple_loss=0.3182, pruned_loss=0.08676, over 8139.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2972, pruned_loss=0.06967, over 1616943.31 frames. ], batch size: 22, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:50:08,290 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0251, 3.8795, 2.2881, 2.7271, 2.7272, 2.0887, 2.8114, 2.8710], + device='cuda:1'), covar=tensor([0.1604, 0.0327, 0.1096, 0.0719, 0.0750, 0.1420, 0.1030, 0.1019], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0238, 0.0332, 0.0307, 0.0305, 0.0332, 0.0348, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:50:31,463 INFO [train.py:901] (1/4) Epoch 16, batch 1400, loss[loss=0.1876, simple_loss=0.2651, pruned_loss=0.05507, over 7421.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2968, pruned_loss=0.06936, over 1617306.44 frames. ], batch size: 17, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:50:34,437 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5336, 2.3598, 4.2631, 1.3565, 3.0069, 2.2382, 1.6733, 2.7916], + device='cuda:1'), covar=tensor([0.1976, 0.2491, 0.0768, 0.4533, 0.1655, 0.3150, 0.2220, 0.2465], + device='cuda:1'), in_proj_covar=tensor([0.0508, 0.0560, 0.0542, 0.0610, 0.0631, 0.0575, 0.0503, 0.0624], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:50:45,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.319e+02 2.799e+02 3.491e+02 7.123e+02, threshold=5.597e+02, percent-clipped=1.0 +2023-02-06 17:50:49,432 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:50:55,487 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122681.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:50:57,007 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122683.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:51:05,788 INFO [train.py:901] (1/4) Epoch 16, batch 1450, loss[loss=0.2396, simple_loss=0.3151, pruned_loss=0.08204, over 8016.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2964, pruned_loss=0.0691, over 1614989.65 frames. ], batch size: 22, lr: 4.81e-03, grad_scale: 4.0 +2023-02-06 17:51:13,476 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 17:51:15,674 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122708.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:51:42,664 INFO [train.py:901] (1/4) Epoch 16, batch 1500, loss[loss=0.1668, simple_loss=0.2423, pruned_loss=0.04562, over 7440.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.296, pruned_loss=0.06871, over 1619430.15 frames. ], batch size: 17, lr: 4.81e-03, grad_scale: 4.0 +2023-02-06 17:51:56,876 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 2.515e+02 3.024e+02 4.111e+02 8.238e+02, threshold=6.047e+02, percent-clipped=9.0 +2023-02-06 17:52:05,879 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8049, 1.5581, 2.0277, 1.7486, 1.9100, 1.7714, 1.5587, 1.1728], + device='cuda:1'), covar=tensor([0.3270, 0.3329, 0.1234, 0.2173, 0.1659, 0.1983, 0.1416, 0.3151], + device='cuda:1'), in_proj_covar=tensor([0.0909, 0.0923, 0.0758, 0.0893, 0.0952, 0.0840, 0.0720, 0.0796], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 17:52:16,427 INFO [train.py:901] (1/4) Epoch 16, batch 1550, loss[loss=0.2202, simple_loss=0.2992, pruned_loss=0.07064, over 8517.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2965, pruned_loss=0.06923, over 1619851.18 frames. ], batch size: 28, lr: 4.81e-03, grad_scale: 4.0 +2023-02-06 17:52:16,637 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122796.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:52:21,765 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 17:52:41,664 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122830.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:52:41,848 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3157, 1.9835, 2.7103, 2.2431, 2.5663, 2.2328, 1.9514, 1.4068], + device='cuda:1'), covar=tensor([0.4271, 0.4324, 0.1537, 0.2890, 0.2146, 0.2525, 0.1716, 0.4506], + device='cuda:1'), in_proj_covar=tensor([0.0907, 0.0921, 0.0758, 0.0893, 0.0949, 0.0838, 0.0717, 0.0794], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 17:52:52,396 INFO [train.py:901] (1/4) Epoch 16, batch 1600, loss[loss=0.196, simple_loss=0.2879, pruned_loss=0.05201, over 8454.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2967, pruned_loss=0.06885, over 1618800.57 frames. ], batch size: 27, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:52:55,444 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122850.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:53:07,650 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.555e+02 3.178e+02 4.067e+02 1.179e+03, threshold=6.355e+02, percent-clipped=12.0 +2023-02-06 17:53:13,373 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:53:23,678 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122890.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:53:27,662 INFO [train.py:901] (1/4) Epoch 16, batch 1650, loss[loss=0.198, simple_loss=0.2716, pruned_loss=0.06216, over 7791.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2961, pruned_loss=0.06865, over 1617413.56 frames. ], batch size: 19, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:53:28,087 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 17:53:28,529 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6293, 1.4038, 1.6398, 1.2084, 0.9382, 1.3129, 1.4396, 1.1994], + device='cuda:1'), covar=tensor([0.0566, 0.1268, 0.1735, 0.1510, 0.0631, 0.1586, 0.0747, 0.0704], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0150, 0.0188, 0.0155, 0.0099, 0.0161, 0.0113, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 17:53:49,589 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122928.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:54:02,346 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122945.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:54:02,856 INFO [train.py:901] (1/4) Epoch 16, batch 1700, loss[loss=0.1996, simple_loss=0.276, pruned_loss=0.06159, over 8140.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2967, pruned_loss=0.06892, over 1617871.00 frames. ], batch size: 22, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:54:08,405 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122953.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:54:11,198 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4167, 1.6408, 1.6524, 0.9791, 1.7025, 1.2667, 0.2465, 1.5694], + device='cuda:1'), covar=tensor([0.0322, 0.0261, 0.0230, 0.0377, 0.0266, 0.0715, 0.0632, 0.0221], + device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0360, 0.0313, 0.0413, 0.0346, 0.0504, 0.0368, 0.0386], + device='cuda:1'), out_proj_covar=tensor([1.1497e-04, 9.6637e-05, 8.3694e-05, 1.1138e-04, 9.3238e-05, 1.4601e-04, + 1.0098e-04, 1.0457e-04], device='cuda:1') +2023-02-06 17:54:17,609 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.342e+02 2.881e+02 3.479e+02 7.679e+02, threshold=5.763e+02, percent-clipped=3.0 +2023-02-06 17:54:38,078 INFO [train.py:901] (1/4) Epoch 16, batch 1750, loss[loss=0.2353, simple_loss=0.3178, pruned_loss=0.07637, over 8364.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2967, pruned_loss=0.06954, over 1611688.15 frames. ], batch size: 24, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:55:12,114 INFO [train.py:901] (1/4) Epoch 16, batch 1800, loss[loss=0.1978, simple_loss=0.2798, pruned_loss=0.05795, over 7812.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2967, pruned_loss=0.06927, over 1611538.95 frames. ], batch size: 19, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:55:16,365 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123052.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:55:25,147 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0344, 3.8158, 2.2755, 2.7576, 2.9532, 2.1788, 2.8733, 2.9604], + device='cuda:1'), covar=tensor([0.1447, 0.0264, 0.1002, 0.0741, 0.0623, 0.1196, 0.0911, 0.1001], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0234, 0.0327, 0.0303, 0.0302, 0.0331, 0.0345, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 17:55:27,703 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 2.489e+02 2.922e+02 3.750e+02 7.056e+02, threshold=5.843e+02, percent-clipped=4.0 +2023-02-06 17:55:35,410 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123077.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:55:48,778 INFO [train.py:901] (1/4) Epoch 16, batch 1850, loss[loss=0.1977, simple_loss=0.2783, pruned_loss=0.05856, over 8239.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2958, pruned_loss=0.06856, over 1613497.44 frames. ], batch size: 22, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:56:22,234 INFO [train.py:901] (1/4) Epoch 16, batch 1900, loss[loss=0.237, simple_loss=0.3301, pruned_loss=0.07193, over 8468.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2964, pruned_loss=0.06895, over 1616842.67 frames. ], batch size: 29, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:56:36,280 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.569e+02 3.077e+02 4.069e+02 9.708e+02, threshold=6.154e+02, percent-clipped=7.0 +2023-02-06 17:56:57,738 INFO [train.py:901] (1/4) Epoch 16, batch 1950, loss[loss=0.2149, simple_loss=0.3038, pruned_loss=0.06301, over 8199.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2986, pruned_loss=0.07032, over 1618585.33 frames. ], batch size: 23, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:56:59,124 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 17:57:01,389 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123201.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:57:11,332 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 17:57:18,801 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123226.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:57:24,080 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=123234.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:57:30,741 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 17:57:32,102 INFO [train.py:901] (1/4) Epoch 16, batch 2000, loss[loss=0.2003, simple_loss=0.2709, pruned_loss=0.06487, over 7535.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2982, pruned_loss=0.07045, over 1613829.36 frames. ], batch size: 18, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:57:46,355 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 2.202e+02 2.631e+02 3.355e+02 6.225e+02, threshold=5.263e+02, percent-clipped=1.0 +2023-02-06 17:58:05,871 INFO [train.py:901] (1/4) Epoch 16, batch 2050, loss[loss=0.2052, simple_loss=0.29, pruned_loss=0.06018, over 8531.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2966, pruned_loss=0.06915, over 1612557.73 frames. ], batch size: 31, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:58:31,864 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123332.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:58:41,130 INFO [train.py:901] (1/4) Epoch 16, batch 2100, loss[loss=0.2003, simple_loss=0.2722, pruned_loss=0.06415, over 7321.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2973, pruned_loss=0.06953, over 1614674.27 frames. ], batch size: 16, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:58:43,354 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:58:54,988 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.773e+02 2.517e+02 3.000e+02 3.631e+02 1.037e+03, threshold=6.000e+02, percent-clipped=6.0 +2023-02-06 17:59:14,280 INFO [train.py:901] (1/4) Epoch 16, batch 2150, loss[loss=0.2085, simple_loss=0.2913, pruned_loss=0.06282, over 8509.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2979, pruned_loss=0.06964, over 1616041.07 frames. ], batch size: 26, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:59:19,800 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2260, 2.0168, 2.7538, 2.2769, 2.7202, 2.2368, 1.9009, 1.2997], + device='cuda:1'), covar=tensor([0.4923, 0.4774, 0.1633, 0.3184, 0.2237, 0.2921, 0.1985, 0.5170], + device='cuda:1'), in_proj_covar=tensor([0.0920, 0.0928, 0.0763, 0.0900, 0.0964, 0.0846, 0.0720, 0.0803], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 17:59:22,900 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1632, 4.1027, 3.8197, 2.6670, 3.7564, 3.7678, 3.9060, 3.4552], + device='cuda:1'), covar=tensor([0.0835, 0.0615, 0.0928, 0.3455, 0.0795, 0.1064, 0.1112, 0.1041], + device='cuda:1'), in_proj_covar=tensor([0.0493, 0.0410, 0.0410, 0.0511, 0.0404, 0.0410, 0.0397, 0.0358], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 17:59:50,134 INFO [train.py:901] (1/4) Epoch 16, batch 2200, loss[loss=0.2063, simple_loss=0.2822, pruned_loss=0.06518, over 7532.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2996, pruned_loss=0.07081, over 1614151.32 frames. ], batch size: 18, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 17:59:50,269 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123446.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:00:04,134 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.316e+02 2.694e+02 3.295e+02 4.036e+02 1.292e+03, threshold=6.590e+02, percent-clipped=6.0 +2023-02-06 18:00:23,381 INFO [train.py:901] (1/4) Epoch 16, batch 2250, loss[loss=0.1773, simple_loss=0.2699, pruned_loss=0.04238, over 8457.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2985, pruned_loss=0.06999, over 1615311.80 frames. ], batch size: 25, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:00:46,373 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 18:00:58,620 INFO [train.py:901] (1/4) Epoch 16, batch 2300, loss[loss=0.2455, simple_loss=0.3141, pruned_loss=0.0884, over 8467.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2986, pruned_loss=0.0703, over 1612156.92 frames. ], batch size: 29, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:01:13,226 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.374e+02 2.935e+02 3.719e+02 2.594e+03, threshold=5.871e+02, percent-clipped=2.0 +2023-02-06 18:01:22,768 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.8955, 1.0921, 1.0542, 0.5647, 1.0795, 0.9026, 0.0834, 1.0449], + device='cuda:1'), covar=tensor([0.0311, 0.0264, 0.0233, 0.0416, 0.0298, 0.0670, 0.0574, 0.0235], + device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0364, 0.0313, 0.0416, 0.0350, 0.0507, 0.0372, 0.0389], + device='cuda:1'), out_proj_covar=tensor([1.1640e-04, 9.7522e-05, 8.3455e-05, 1.1214e-04, 9.4408e-05, 1.4701e-04, + 1.0201e-04, 1.0529e-04], device='cuda:1') +2023-02-06 18:01:32,627 INFO [train.py:901] (1/4) Epoch 16, batch 2350, loss[loss=0.3289, simple_loss=0.3846, pruned_loss=0.1366, over 7300.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2988, pruned_loss=0.0705, over 1613449.44 frames. ], batch size: 72, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:01:38,852 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:01:55,681 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:02:06,124 INFO [train.py:901] (1/4) Epoch 16, batch 2400, loss[loss=0.1582, simple_loss=0.2381, pruned_loss=0.03916, over 7535.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2979, pruned_loss=0.07037, over 1611837.66 frames. ], batch size: 18, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:02:22,334 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.321e+02 3.011e+02 3.485e+02 7.740e+02, threshold=6.021e+02, percent-clipped=5.0 +2023-02-06 18:02:28,432 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=123676.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:02:42,472 INFO [train.py:901] (1/4) Epoch 16, batch 2450, loss[loss=0.2167, simple_loss=0.2852, pruned_loss=0.0741, over 6836.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2973, pruned_loss=0.06998, over 1609430.16 frames. ], batch size: 15, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:03:15,350 INFO [train.py:901] (1/4) Epoch 16, batch 2500, loss[loss=0.2368, simple_loss=0.3172, pruned_loss=0.07819, over 8360.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2962, pruned_loss=0.06971, over 1610266.29 frames. ], batch size: 24, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:03:25,449 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8191, 1.6896, 1.9705, 1.6020, 1.1732, 1.8085, 2.0899, 1.9574], + device='cuda:1'), covar=tensor([0.0481, 0.1141, 0.1529, 0.1375, 0.0609, 0.1325, 0.0664, 0.0568], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0190, 0.0157, 0.0100, 0.0163, 0.0114, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 18:03:29,366 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 2.388e+02 3.009e+02 3.987e+02 1.163e+03, threshold=6.019e+02, percent-clipped=7.0 +2023-02-06 18:03:46,949 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=123790.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:03:47,737 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123791.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:03:50,918 INFO [train.py:901] (1/4) Epoch 16, batch 2550, loss[loss=0.1834, simple_loss=0.2739, pruned_loss=0.04646, over 8029.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2968, pruned_loss=0.06981, over 1609487.09 frames. ], batch size: 22, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:04:24,937 INFO [train.py:901] (1/4) Epoch 16, batch 2600, loss[loss=0.1992, simple_loss=0.2918, pruned_loss=0.05329, over 8465.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2955, pruned_loss=0.06874, over 1612216.70 frames. ], batch size: 25, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:04:38,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.447e+02 2.814e+02 3.524e+02 5.517e+02, threshold=5.629e+02, percent-clipped=0.0 +2023-02-06 18:04:54,348 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123890.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:04:58,762 INFO [train.py:901] (1/4) Epoch 16, batch 2650, loss[loss=0.2301, simple_loss=0.3015, pruned_loss=0.07934, over 7663.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2965, pruned_loss=0.06924, over 1612073.53 frames. ], batch size: 19, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:05:06,338 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123905.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:05:34,184 INFO [train.py:901] (1/4) Epoch 16, batch 2700, loss[loss=0.2268, simple_loss=0.3061, pruned_loss=0.07371, over 8292.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2972, pruned_loss=0.06968, over 1614115.60 frames. ], batch size: 23, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:05:48,215 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.455e+02 3.188e+02 4.135e+02 8.908e+02, threshold=6.377e+02, percent-clipped=7.0 +2023-02-06 18:06:07,604 INFO [train.py:901] (1/4) Epoch 16, batch 2750, loss[loss=0.1595, simple_loss=0.2406, pruned_loss=0.03922, over 7795.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2958, pruned_loss=0.06905, over 1612018.36 frames. ], batch size: 19, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:06:45,100 INFO [train.py:901] (1/4) Epoch 16, batch 2800, loss[loss=0.1985, simple_loss=0.2754, pruned_loss=0.06081, over 8137.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2951, pruned_loss=0.06832, over 1611738.89 frames. ], batch size: 22, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:06:46,030 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124047.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:06:50,855 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124054.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:06:59,430 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.578e+02 3.039e+02 4.001e+02 1.196e+03, threshold=6.079e+02, percent-clipped=5.0 +2023-02-06 18:07:03,053 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124072.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:07:08,255 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:07:18,993 INFO [train.py:901] (1/4) Epoch 16, batch 2850, loss[loss=0.1905, simple_loss=0.2801, pruned_loss=0.0505, over 8355.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2968, pruned_loss=0.06906, over 1616402.67 frames. ], batch size: 24, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:07:22,873 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 18:07:55,341 INFO [train.py:901] (1/4) Epoch 16, batch 2900, loss[loss=0.1902, simple_loss=0.2794, pruned_loss=0.05057, over 8358.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2966, pruned_loss=0.06898, over 1607310.91 frames. ], batch size: 24, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:08:06,263 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:08:06,299 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5562, 2.3125, 3.2074, 2.5719, 2.9324, 2.5148, 2.0987, 1.7709], + device='cuda:1'), covar=tensor([0.4712, 0.4853, 0.1613, 0.3263, 0.2690, 0.2647, 0.1864, 0.5206], + device='cuda:1'), in_proj_covar=tensor([0.0914, 0.0924, 0.0758, 0.0894, 0.0961, 0.0848, 0.0723, 0.0797], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:08:10,037 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.456e+02 3.206e+02 4.387e+02 8.191e+02, threshold=6.412e+02, percent-clipped=4.0 +2023-02-06 18:08:22,881 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124186.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:08:29,509 INFO [train.py:901] (1/4) Epoch 16, batch 2950, loss[loss=0.1943, simple_loss=0.2652, pruned_loss=0.06164, over 7542.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2976, pruned_loss=0.06986, over 1607666.58 frames. ], batch size: 18, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:08:29,742 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0376, 2.4639, 2.7931, 1.4929, 3.0306, 1.7616, 1.3914, 1.9053], + device='cuda:1'), covar=tensor([0.0678, 0.0321, 0.0203, 0.0636, 0.0327, 0.0681, 0.0800, 0.0486], + device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0363, 0.0311, 0.0417, 0.0350, 0.0508, 0.0371, 0.0387], + device='cuda:1'), out_proj_covar=tensor([1.1693e-04, 9.7248e-05, 8.2672e-05, 1.1219e-04, 9.4535e-05, 1.4726e-04, + 1.0189e-04, 1.0494e-04], device='cuda:1') +2023-02-06 18:08:35,666 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 18:08:55,342 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124234.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:09:01,454 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.69 vs. limit=5.0 +2023-02-06 18:09:03,840 INFO [train.py:901] (1/4) Epoch 16, batch 3000, loss[loss=0.2278, simple_loss=0.311, pruned_loss=0.07227, over 8495.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2979, pruned_loss=0.07009, over 1608773.49 frames. ], batch size: 28, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:09:03,840 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 18:09:16,275 INFO [train.py:935] (1/4) Epoch 16, validation: loss=0.1794, simple_loss=0.2796, pruned_loss=0.03958, over 944034.00 frames. +2023-02-06 18:09:16,277 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 18:09:32,708 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.393e+02 2.939e+02 3.627e+02 1.404e+03, threshold=5.877e+02, percent-clipped=2.0 +2023-02-06 18:09:49,103 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124290.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:09:52,919 INFO [train.py:901] (1/4) Epoch 16, batch 3050, loss[loss=0.2025, simple_loss=0.2834, pruned_loss=0.06081, over 7292.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.298, pruned_loss=0.06975, over 1613398.37 frames. ], batch size: 16, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:10:25,965 INFO [train.py:901] (1/4) Epoch 16, batch 3100, loss[loss=0.2554, simple_loss=0.3324, pruned_loss=0.08917, over 8342.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2981, pruned_loss=0.07004, over 1613697.46 frames. ], batch size: 26, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:10:28,059 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:10:39,312 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124366.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:10:39,810 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.464e+02 2.975e+02 4.095e+02 1.383e+03, threshold=5.950e+02, percent-clipped=6.0 +2023-02-06 18:11:01,474 INFO [train.py:901] (1/4) Epoch 16, batch 3150, loss[loss=0.2241, simple_loss=0.307, pruned_loss=0.0706, over 8318.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2986, pruned_loss=0.0703, over 1617433.22 frames. ], batch size: 25, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:11:02,941 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:11:21,450 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124424.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:11:36,587 INFO [train.py:901] (1/4) Epoch 16, batch 3200, loss[loss=0.2045, simple_loss=0.2787, pruned_loss=0.06512, over 7705.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2972, pruned_loss=0.06973, over 1616303.04 frames. ], batch size: 18, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:11:39,401 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124450.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:11:50,425 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.712e+02 2.552e+02 3.102e+02 3.772e+02 6.284e+02, threshold=6.205e+02, percent-clipped=3.0 +2023-02-06 18:12:09,966 INFO [train.py:901] (1/4) Epoch 16, batch 3250, loss[loss=0.2221, simple_loss=0.3101, pruned_loss=0.06704, over 8464.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2967, pruned_loss=0.06953, over 1615083.35 frames. ], batch size: 25, lr: 4.77e-03, grad_scale: 8.0 +2023-02-06 18:12:23,026 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124513.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:12:40,935 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124539.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:12:45,404 INFO [train.py:901] (1/4) Epoch 16, batch 3300, loss[loss=0.217, simple_loss=0.2978, pruned_loss=0.06812, over 8139.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2974, pruned_loss=0.06958, over 1613426.36 frames. ], batch size: 22, lr: 4.77e-03, grad_scale: 8.0 +2023-02-06 18:12:59,420 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.425e+02 2.919e+02 3.659e+02 6.879e+02, threshold=5.837e+02, percent-clipped=1.0 +2023-02-06 18:13:18,842 INFO [train.py:901] (1/4) Epoch 16, batch 3350, loss[loss=0.1817, simple_loss=0.2676, pruned_loss=0.0479, over 8245.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2962, pruned_loss=0.06875, over 1611873.70 frames. ], batch size: 22, lr: 4.77e-03, grad_scale: 8.0 +2023-02-06 18:13:25,351 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:13:40,971 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 +2023-02-06 18:13:43,479 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:13:46,020 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:13:54,715 INFO [train.py:901] (1/4) Epoch 16, batch 3400, loss[loss=0.1961, simple_loss=0.2849, pruned_loss=0.05361, over 8554.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2964, pruned_loss=0.06846, over 1613630.23 frames. ], batch size: 34, lr: 4.77e-03, grad_scale: 8.0 +2023-02-06 18:14:01,601 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124656.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:14:08,833 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.420e+02 3.011e+02 3.525e+02 7.222e+02, threshold=6.022e+02, percent-clipped=3.0 +2023-02-06 18:14:18,588 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124681.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:14:28,866 INFO [train.py:901] (1/4) Epoch 16, batch 3450, loss[loss=0.1908, simple_loss=0.2782, pruned_loss=0.05167, over 7935.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.297, pruned_loss=0.06875, over 1613773.32 frames. ], batch size: 20, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:14:38,558 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124710.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:14:48,196 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7185, 5.7847, 5.0728, 2.4876, 5.1049, 5.7031, 5.3321, 5.1774], + device='cuda:1'), covar=tensor([0.0471, 0.0405, 0.0806, 0.4187, 0.0667, 0.0631, 0.0918, 0.0535], + device='cuda:1'), in_proj_covar=tensor([0.0489, 0.0408, 0.0410, 0.0511, 0.0401, 0.0408, 0.0396, 0.0355], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 18:15:04,146 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3221, 1.5621, 1.6366, 0.9905, 1.6876, 1.2214, 0.3140, 1.5009], + device='cuda:1'), covar=tensor([0.0408, 0.0313, 0.0232, 0.0439, 0.0336, 0.0829, 0.0691, 0.0234], + device='cuda:1'), in_proj_covar=tensor([0.0424, 0.0363, 0.0310, 0.0415, 0.0351, 0.0508, 0.0371, 0.0387], + device='cuda:1'), out_proj_covar=tensor([1.1649e-04, 9.7225e-05, 8.2362e-05, 1.1162e-04, 9.4818e-05, 1.4724e-04, + 1.0156e-04, 1.0472e-04], device='cuda:1') +2023-02-06 18:15:05,363 INFO [train.py:901] (1/4) Epoch 16, batch 3500, loss[loss=0.2188, simple_loss=0.3012, pruned_loss=0.06816, over 8453.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2951, pruned_loss=0.06753, over 1613971.47 frames. ], batch size: 27, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:15:07,658 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124749.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:09,711 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8730, 5.9730, 5.2610, 2.3176, 5.3410, 5.6856, 5.5419, 5.2486], + device='cuda:1'), covar=tensor([0.0509, 0.0414, 0.0889, 0.4335, 0.0684, 0.0633, 0.0963, 0.0520], + device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0406, 0.0409, 0.0509, 0.0400, 0.0407, 0.0395, 0.0354], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 18:15:20,551 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.534e+02 3.082e+02 3.894e+02 7.146e+02, threshold=6.164e+02, percent-clipped=3.0 +2023-02-06 18:15:22,114 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124769.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:38,223 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 18:15:38,977 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:39,096 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:39,719 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124795.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:40,182 INFO [train.py:901] (1/4) Epoch 16, batch 3550, loss[loss=0.1818, simple_loss=0.266, pruned_loss=0.04875, over 8133.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2944, pruned_loss=0.06727, over 1616629.25 frames. ], batch size: 22, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:15:56,878 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124820.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:16:00,149 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124825.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:16:04,090 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8427, 1.5772, 6.0108, 2.0220, 5.3413, 5.1682, 5.5754, 5.4132], + device='cuda:1'), covar=tensor([0.0456, 0.4666, 0.0366, 0.3799, 0.1001, 0.0798, 0.0452, 0.0510], + device='cuda:1'), in_proj_covar=tensor([0.0561, 0.0613, 0.0637, 0.0589, 0.0662, 0.0569, 0.0561, 0.0625], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:16:08,893 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124838.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:16:14,226 INFO [train.py:901] (1/4) Epoch 16, batch 3600, loss[loss=0.1881, simple_loss=0.2583, pruned_loss=0.05898, over 7804.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2944, pruned_loss=0.06766, over 1614878.73 frames. ], batch size: 19, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:16:30,802 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.338e+02 2.977e+02 3.463e+02 8.977e+02, threshold=5.954e+02, percent-clipped=2.0 +2023-02-06 18:16:50,928 INFO [train.py:901] (1/4) Epoch 16, batch 3650, loss[loss=0.2273, simple_loss=0.2954, pruned_loss=0.07964, over 7923.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2961, pruned_loss=0.06876, over 1616284.64 frames. ], batch size: 20, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:16:59,893 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124909.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:17:24,965 INFO [train.py:901] (1/4) Epoch 16, batch 3700, loss[loss=0.2613, simple_loss=0.3341, pruned_loss=0.09432, over 6995.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2966, pruned_loss=0.06937, over 1615154.85 frames. ], batch size: 71, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:17:38,858 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 18:17:40,140 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.880e+02 2.643e+02 3.299e+02 4.315e+02 1.525e+03, threshold=6.598e+02, percent-clipped=10.0 +2023-02-06 18:18:01,619 INFO [train.py:901] (1/4) Epoch 16, batch 3750, loss[loss=0.2152, simple_loss=0.301, pruned_loss=0.06467, over 8350.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2951, pruned_loss=0.06814, over 1612423.66 frames. ], batch size: 24, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:18:04,453 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125000.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:18:07,882 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:18:20,985 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125025.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:18:24,558 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125030.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:18:35,169 INFO [train.py:901] (1/4) Epoch 16, batch 3800, loss[loss=0.198, simple_loss=0.2892, pruned_loss=0.05347, over 7806.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2966, pruned_loss=0.06903, over 1615985.55 frames. ], batch size: 20, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:18:49,284 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.284e+02 2.854e+02 3.651e+02 7.015e+02, threshold=5.709e+02, percent-clipped=3.0 +2023-02-06 18:18:58,957 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125081.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:19:10,734 INFO [train.py:901] (1/4) Epoch 16, batch 3850, loss[loss=0.2266, simple_loss=0.311, pruned_loss=0.07116, over 8125.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2956, pruned_loss=0.06879, over 1609903.64 frames. ], batch size: 22, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:19:18,418 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125106.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:19:24,394 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125115.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:19:41,022 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:19:45,083 INFO [train.py:901] (1/4) Epoch 16, batch 3900, loss[loss=0.2192, simple_loss=0.3102, pruned_loss=0.06409, over 8189.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2952, pruned_loss=0.06878, over 1608734.17 frames. ], batch size: 23, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:19:45,101 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 18:19:45,238 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125146.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:19:58,185 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:19:59,305 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 2.507e+02 2.888e+02 3.601e+02 7.393e+02, threshold=5.777e+02, percent-clipped=3.0 +2023-02-06 18:20:09,607 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125182.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:20:15,217 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125190.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:20:19,100 INFO [train.py:901] (1/4) Epoch 16, batch 3950, loss[loss=0.2223, simple_loss=0.2995, pruned_loss=0.0725, over 8283.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2942, pruned_loss=0.06798, over 1608820.09 frames. ], batch size: 23, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:20:55,593 INFO [train.py:901] (1/4) Epoch 16, batch 4000, loss[loss=0.2071, simple_loss=0.2819, pruned_loss=0.0661, over 7719.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2962, pruned_loss=0.06927, over 1606547.80 frames. ], batch size: 18, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:21:08,104 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0918, 1.8783, 1.9135, 1.7617, 1.2997, 1.8785, 1.9793, 1.9224], + device='cuda:1'), covar=tensor([0.0549, 0.0859, 0.1262, 0.1061, 0.0599, 0.1084, 0.0652, 0.0470], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0156, 0.0100, 0.0162, 0.0113, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 18:21:09,913 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 2.424e+02 2.747e+02 3.530e+02 7.172e+02, threshold=5.495e+02, percent-clipped=3.0 +2023-02-06 18:21:22,772 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3852, 1.4250, 1.3068, 1.8190, 0.6688, 1.2033, 1.1883, 1.4410], + device='cuda:1'), covar=tensor([0.0819, 0.0781, 0.1084, 0.0484, 0.1199, 0.1429, 0.0890, 0.0730], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0203, 0.0251, 0.0212, 0.0210, 0.0248, 0.0256, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 18:21:29,135 INFO [train.py:901] (1/4) Epoch 16, batch 4050, loss[loss=0.2204, simple_loss=0.3149, pruned_loss=0.06297, over 8357.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2979, pruned_loss=0.06985, over 1606891.55 frames. ], batch size: 24, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:21:29,950 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125297.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:21:41,126 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4670, 4.4603, 3.9716, 1.9904, 3.9806, 4.0492, 4.0316, 3.8018], + device='cuda:1'), covar=tensor([0.0710, 0.0552, 0.1201, 0.4741, 0.0823, 0.1014, 0.1204, 0.0827], + device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0412, 0.0410, 0.0516, 0.0405, 0.0411, 0.0402, 0.0358], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 18:22:05,146 INFO [train.py:901] (1/4) Epoch 16, batch 4100, loss[loss=0.226, simple_loss=0.3131, pruned_loss=0.06944, over 8463.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.299, pruned_loss=0.07026, over 1611358.16 frames. ], batch size: 39, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:22:19,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.458e+02 2.941e+02 3.398e+02 7.943e+02, threshold=5.881e+02, percent-clipped=6.0 +2023-02-06 18:22:22,354 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125371.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:24,211 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125374.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:37,725 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125394.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:38,928 INFO [train.py:901] (1/4) Epoch 16, batch 4150, loss[loss=0.2091, simple_loss=0.3076, pruned_loss=0.05529, over 8196.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3001, pruned_loss=0.07094, over 1614292.86 frames. ], batch size: 23, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:22:39,124 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125396.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:39,142 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125396.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:42,588 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1532, 2.8078, 3.8616, 1.9460, 2.0585, 3.9736, 1.0174, 2.4391], + device='cuda:1'), covar=tensor([0.1683, 0.1362, 0.0216, 0.2440, 0.3053, 0.0231, 0.2543, 0.1299], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0179, 0.0112, 0.0212, 0.0255, 0.0116, 0.0162, 0.0175], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 18:22:43,282 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9117, 2.3901, 2.5336, 1.3982, 2.5385, 1.5612, 1.4860, 1.9803], + device='cuda:1'), covar=tensor([0.0796, 0.0385, 0.0249, 0.0742, 0.0507, 0.0782, 0.0764, 0.0462], + device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0362, 0.0311, 0.0419, 0.0354, 0.0510, 0.0372, 0.0392], + device='cuda:1'), out_proj_covar=tensor([1.1741e-04, 9.6606e-05, 8.2775e-05, 1.1276e-04, 9.5365e-05, 1.4761e-04, + 1.0191e-04, 1.0613e-04], device='cuda:1') +2023-02-06 18:22:55,759 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:23:09,065 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125439.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:23:14,194 INFO [train.py:901] (1/4) Epoch 16, batch 4200, loss[loss=0.1977, simple_loss=0.2841, pruned_loss=0.05568, over 8346.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2982, pruned_loss=0.07035, over 1611626.21 frames. ], batch size: 24, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:23:29,128 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.508e+02 2.881e+02 3.373e+02 7.881e+02, threshold=5.761e+02, percent-clipped=2.0 +2023-02-06 18:23:39,903 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 18:23:42,027 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7029, 1.3613, 1.6084, 1.1591, 0.9144, 1.4109, 1.5039, 1.5772], + device='cuda:1'), covar=tensor([0.0478, 0.1247, 0.1621, 0.1472, 0.0588, 0.1477, 0.0677, 0.0567], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0189, 0.0156, 0.0100, 0.0162, 0.0113, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 18:23:44,591 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:23:48,645 INFO [train.py:901] (1/4) Epoch 16, batch 4250, loss[loss=0.2184, simple_loss=0.2918, pruned_loss=0.07247, over 8603.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2969, pruned_loss=0.06929, over 1611047.45 frames. ], batch size: 31, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:24:01,592 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 18:24:09,734 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4704, 2.0370, 2.8511, 2.2934, 2.7010, 2.4079, 2.1340, 1.5507], + device='cuda:1'), covar=tensor([0.4543, 0.4490, 0.1662, 0.3334, 0.2361, 0.2551, 0.1653, 0.4914], + device='cuda:1'), in_proj_covar=tensor([0.0913, 0.0921, 0.0758, 0.0898, 0.0961, 0.0849, 0.0720, 0.0795], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:24:11,893 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 18:24:23,120 INFO [train.py:901] (1/4) Epoch 16, batch 4300, loss[loss=0.1944, simple_loss=0.2731, pruned_loss=0.05779, over 8241.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2982, pruned_loss=0.06989, over 1617132.08 frames. ], batch size: 22, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:24:28,629 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125553.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:24:38,395 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.516e+02 3.115e+02 4.119e+02 8.810e+02, threshold=6.231e+02, percent-clipped=6.0 +2023-02-06 18:24:46,751 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125578.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:24:52,791 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6455, 2.5188, 1.8335, 2.1945, 2.1464, 1.5467, 2.0182, 2.0437], + device='cuda:1'), covar=tensor([0.1266, 0.0298, 0.1029, 0.0609, 0.0635, 0.1348, 0.0948, 0.0966], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0232, 0.0327, 0.0304, 0.0300, 0.0333, 0.0346, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 18:24:58,867 INFO [train.py:901] (1/4) Epoch 16, batch 4350, loss[loss=0.2052, simple_loss=0.2908, pruned_loss=0.0598, over 8246.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2971, pruned_loss=0.06879, over 1616601.58 frames. ], batch size: 24, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:25:02,856 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 18:25:05,371 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:25:16,541 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125621.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:25:33,361 INFO [train.py:901] (1/4) Epoch 16, batch 4400, loss[loss=0.2043, simple_loss=0.2941, pruned_loss=0.05723, over 8292.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2964, pruned_loss=0.06862, over 1615451.48 frames. ], batch size: 23, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:25:34,038 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 18:25:48,659 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.489e+02 3.156e+02 3.927e+02 6.760e+02, threshold=6.312e+02, percent-clipped=2.0 +2023-02-06 18:26:09,565 INFO [train.py:901] (1/4) Epoch 16, batch 4450, loss[loss=0.2381, simple_loss=0.3042, pruned_loss=0.08597, over 7961.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2962, pruned_loss=0.06902, over 1611666.06 frames. ], batch size: 21, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:26:14,203 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 18:26:23,713 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6212, 1.8979, 2.0246, 1.3726, 2.1618, 1.4214, 0.5451, 1.8405], + device='cuda:1'), covar=tensor([0.0516, 0.0280, 0.0206, 0.0447, 0.0327, 0.0826, 0.0754, 0.0242], + device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0361, 0.0314, 0.0420, 0.0354, 0.0511, 0.0372, 0.0391], + device='cuda:1'), out_proj_covar=tensor([1.1825e-04, 9.6388e-05, 8.3468e-05, 1.1270e-04, 9.5577e-05, 1.4802e-04, + 1.0204e-04, 1.0591e-04], device='cuda:1') +2023-02-06 18:26:24,207 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:26:38,219 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125738.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:26:41,747 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125743.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:26:43,640 INFO [train.py:901] (1/4) Epoch 16, batch 4500, loss[loss=0.2438, simple_loss=0.3099, pruned_loss=0.08883, over 7934.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2965, pruned_loss=0.06914, over 1611121.51 frames. ], batch size: 20, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:26:57,818 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.361e+02 2.740e+02 3.373e+02 6.169e+02, threshold=5.479e+02, percent-clipped=0.0 +2023-02-06 18:27:04,102 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 18:27:10,713 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125783.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:27:19,278 INFO [train.py:901] (1/4) Epoch 16, batch 4550, loss[loss=0.1662, simple_loss=0.2472, pruned_loss=0.04265, over 7656.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2966, pruned_loss=0.06905, over 1610150.00 frames. ], batch size: 19, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:27:20,086 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8645, 1.4029, 6.0028, 2.1129, 5.4341, 4.9930, 5.5204, 5.4076], + device='cuda:1'), covar=tensor([0.0426, 0.4692, 0.0361, 0.3661, 0.0938, 0.0889, 0.0465, 0.0435], + device='cuda:1'), in_proj_covar=tensor([0.0560, 0.0619, 0.0641, 0.0590, 0.0660, 0.0573, 0.0563, 0.0630], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:27:45,705 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:27:54,456 INFO [train.py:901] (1/4) Epoch 16, batch 4600, loss[loss=0.1891, simple_loss=0.2678, pruned_loss=0.05516, over 8298.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2968, pruned_loss=0.0691, over 1610442.63 frames. ], batch size: 23, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:27:59,460 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125853.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:28:03,905 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.89 vs. limit=5.0 +2023-02-06 18:28:05,126 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125861.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:28:08,980 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.490e+02 3.040e+02 3.897e+02 1.241e+03, threshold=6.080e+02, percent-clipped=8.0 +2023-02-06 18:28:22,168 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:28:30,021 INFO [train.py:901] (1/4) Epoch 16, batch 4650, loss[loss=0.2362, simple_loss=0.3264, pruned_loss=0.07301, over 8029.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2963, pruned_loss=0.06867, over 1611598.21 frames. ], batch size: 22, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:28:31,586 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125898.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:29:06,101 INFO [train.py:901] (1/4) Epoch 16, batch 4700, loss[loss=0.2103, simple_loss=0.2845, pruned_loss=0.06806, over 7807.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2965, pruned_loss=0.06887, over 1610402.95 frames. ], batch size: 20, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:29:18,971 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:29:20,231 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.580e+02 3.138e+02 4.127e+02 1.212e+03, threshold=6.277e+02, percent-clipped=5.0 +2023-02-06 18:29:34,299 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-02-06 18:29:39,519 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 18:29:39,834 INFO [train.py:901] (1/4) Epoch 16, batch 4750, loss[loss=0.1838, simple_loss=0.281, pruned_loss=0.04326, over 8558.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2964, pruned_loss=0.06868, over 1609858.25 frames. ], batch size: 31, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:29:55,976 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126016.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:30:11,192 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 18:30:13,728 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 18:30:15,807 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1961, 1.6657, 3.3940, 1.4692, 2.3066, 3.7190, 3.7532, 3.2519], + device='cuda:1'), covar=tensor([0.0953, 0.1612, 0.0335, 0.2068, 0.1089, 0.0226, 0.0510, 0.0538], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0310, 0.0275, 0.0300, 0.0292, 0.0250, 0.0384, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 18:30:16,331 INFO [train.py:901] (1/4) Epoch 16, batch 4800, loss[loss=0.1998, simple_loss=0.2872, pruned_loss=0.05624, over 8344.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2955, pruned_loss=0.06822, over 1608473.74 frames. ], batch size: 26, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:30:31,318 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.301e+02 2.788e+02 3.330e+02 6.705e+02, threshold=5.575e+02, percent-clipped=2.0 +2023-02-06 18:30:37,059 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3903, 1.5759, 1.3456, 2.0836, 1.0472, 1.1649, 1.4812, 1.7023], + device='cuda:1'), covar=tensor([0.0827, 0.0782, 0.0911, 0.0366, 0.0897, 0.1288, 0.0751, 0.0672], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0204, 0.0250, 0.0214, 0.0211, 0.0249, 0.0257, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 18:30:40,380 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:30:45,002 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=126087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:30:46,461 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126089.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:30:51,047 INFO [train.py:901] (1/4) Epoch 16, batch 4850, loss[loss=0.1951, simple_loss=0.287, pruned_loss=0.05159, over 8246.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2958, pruned_loss=0.0685, over 1605749.23 frames. ], batch size: 24, lr: 4.74e-03, grad_scale: 16.0 +2023-02-06 18:30:59,315 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6626, 1.9742, 2.0775, 1.2874, 2.2513, 1.3993, 0.6425, 1.9120], + device='cuda:1'), covar=tensor([0.0526, 0.0313, 0.0236, 0.0549, 0.0318, 0.0905, 0.0750, 0.0258], + device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0362, 0.0314, 0.0421, 0.0353, 0.0511, 0.0371, 0.0390], + device='cuda:1'), out_proj_covar=tensor([1.1717e-04, 9.6436e-05, 8.3497e-05, 1.1313e-04, 9.5198e-05, 1.4780e-04, + 1.0152e-04, 1.0561e-04], device='cuda:1') +2023-02-06 18:30:59,983 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126109.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:31:01,795 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 18:31:03,310 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126114.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:31:19,043 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:31:26,987 INFO [train.py:901] (1/4) Epoch 16, batch 4900, loss[loss=0.1828, simple_loss=0.2589, pruned_loss=0.05336, over 7209.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2947, pruned_loss=0.06784, over 1604750.32 frames. ], batch size: 16, lr: 4.74e-03, grad_scale: 16.0 +2023-02-06 18:31:32,564 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126154.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:31:41,753 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.384e+02 3.140e+02 3.836e+02 7.587e+02, threshold=6.281e+02, percent-clipped=5.0 +2023-02-06 18:31:50,102 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126179.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:32:01,584 INFO [train.py:901] (1/4) Epoch 16, batch 4950, loss[loss=0.1773, simple_loss=0.2627, pruned_loss=0.04597, over 8199.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2947, pruned_loss=0.06831, over 1606207.69 frames. ], batch size: 23, lr: 4.74e-03, grad_scale: 16.0 +2023-02-06 18:32:06,034 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126202.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:32:35,840 INFO [train.py:901] (1/4) Epoch 16, batch 5000, loss[loss=0.1994, simple_loss=0.2779, pruned_loss=0.06042, over 8475.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2955, pruned_loss=0.06796, over 1610571.17 frames. ], batch size: 25, lr: 4.74e-03, grad_scale: 16.0 +2023-02-06 18:32:37,272 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0913, 1.5449, 4.3243, 1.6159, 3.7793, 3.5958, 3.8806, 3.7422], + device='cuda:1'), covar=tensor([0.0637, 0.4204, 0.0575, 0.3901, 0.1293, 0.1024, 0.0621, 0.0757], + device='cuda:1'), in_proj_covar=tensor([0.0564, 0.0622, 0.0647, 0.0593, 0.0669, 0.0576, 0.0565, 0.0633], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:32:50,294 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.421e+02 2.802e+02 3.540e+02 7.456e+02, threshold=5.603e+02, percent-clipped=2.0 +2023-02-06 18:33:10,452 INFO [train.py:901] (1/4) Epoch 16, batch 5050, loss[loss=0.2166, simple_loss=0.3008, pruned_loss=0.06618, over 8030.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2952, pruned_loss=0.06758, over 1614579.07 frames. ], batch size: 22, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:33:38,223 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:33:39,107 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 18:33:41,489 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 18:33:44,853 INFO [train.py:901] (1/4) Epoch 16, batch 5100, loss[loss=0.2552, simple_loss=0.3081, pruned_loss=0.1012, over 7931.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2964, pruned_loss=0.06835, over 1615928.33 frames. ], batch size: 20, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:33:55,147 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=126360.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:33:55,980 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126361.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:34:01,121 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.574e+02 2.967e+02 3.773e+02 8.448e+02, threshold=5.934e+02, percent-clipped=7.0 +2023-02-06 18:34:20,681 INFO [train.py:901] (1/4) Epoch 16, batch 5150, loss[loss=0.2404, simple_loss=0.3109, pruned_loss=0.08495, over 8457.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2963, pruned_loss=0.06884, over 1615703.91 frames. ], batch size: 27, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:34:22,771 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:34:54,884 INFO [train.py:901] (1/4) Epoch 16, batch 5200, loss[loss=0.2546, simple_loss=0.3255, pruned_loss=0.09181, over 8478.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.296, pruned_loss=0.06879, over 1614410.81 frames. ], batch size: 28, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:35:03,396 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126458.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:35:10,025 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.450e+02 2.961e+02 4.009e+02 9.502e+02, threshold=5.923e+02, percent-clipped=8.0 +2023-02-06 18:35:10,198 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126468.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:35:15,126 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126475.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:35:21,833 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126483.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:35:30,966 INFO [train.py:901] (1/4) Epoch 16, batch 5250, loss[loss=0.2909, simple_loss=0.3467, pruned_loss=0.1176, over 8525.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2957, pruned_loss=0.06854, over 1613800.25 frames. ], batch size: 28, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:35:35,230 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5575, 1.9893, 3.3274, 1.3875, 2.4450, 1.9561, 1.7300, 2.4449], + device='cuda:1'), covar=tensor([0.1765, 0.2392, 0.0755, 0.4134, 0.1676, 0.2887, 0.1953, 0.2143], + device='cuda:1'), in_proj_covar=tensor([0.0500, 0.0553, 0.0540, 0.0610, 0.0627, 0.0566, 0.0496, 0.0615], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 18:35:39,844 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 18:36:05,601 INFO [train.py:901] (1/4) Epoch 16, batch 5300, loss[loss=0.2113, simple_loss=0.2934, pruned_loss=0.06458, over 8581.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2965, pruned_loss=0.0694, over 1612624.48 frames. ], batch size: 31, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:36:20,899 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.415e+02 2.951e+02 3.953e+02 1.148e+03, threshold=5.902e+02, percent-clipped=4.0 +2023-02-06 18:36:34,266 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7213, 1.5062, 1.8943, 1.4834, 0.9225, 1.7164, 2.1286, 2.0235], + device='cuda:1'), covar=tensor([0.0483, 0.1248, 0.1631, 0.1446, 0.0607, 0.1399, 0.0655, 0.0562], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0191, 0.0157, 0.0100, 0.0162, 0.0114, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 18:36:41,568 INFO [train.py:901] (1/4) Epoch 16, batch 5350, loss[loss=0.2118, simple_loss=0.2989, pruned_loss=0.06231, over 8105.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2971, pruned_loss=0.0696, over 1612619.64 frames. ], batch size: 23, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:36:50,006 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7963, 1.6494, 2.8003, 1.4000, 2.2117, 3.0192, 3.0946, 2.6123], + device='cuda:1'), covar=tensor([0.1012, 0.1407, 0.0451, 0.1977, 0.1010, 0.0285, 0.0613, 0.0534], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0311, 0.0276, 0.0299, 0.0292, 0.0250, 0.0384, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 18:36:55,927 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6997, 2.0518, 3.1739, 1.5348, 2.5735, 2.0829, 1.7813, 2.4630], + device='cuda:1'), covar=tensor([0.1661, 0.2057, 0.0763, 0.3834, 0.1477, 0.2577, 0.1845, 0.1895], + device='cuda:1'), in_proj_covar=tensor([0.0500, 0.0553, 0.0539, 0.0610, 0.0626, 0.0565, 0.0497, 0.0616], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 18:37:08,154 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5910, 2.1030, 3.2498, 1.4635, 1.4107, 3.0969, 0.7731, 1.9678], + device='cuda:1'), covar=tensor([0.1953, 0.1470, 0.0302, 0.3096, 0.3669, 0.0352, 0.2843, 0.2066], + device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0183, 0.0114, 0.0214, 0.0260, 0.0119, 0.0166, 0.0179], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 18:37:16,915 INFO [train.py:901] (1/4) Epoch 16, batch 5400, loss[loss=0.2402, simple_loss=0.3166, pruned_loss=0.08189, over 8463.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2976, pruned_loss=0.07009, over 1614968.89 frames. ], batch size: 25, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:37:32,198 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.767e+02 2.413e+02 2.875e+02 3.758e+02 9.843e+02, threshold=5.751e+02, percent-clipped=6.0 +2023-02-06 18:37:48,665 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-02-06 18:37:51,443 INFO [train.py:901] (1/4) Epoch 16, batch 5450, loss[loss=0.2158, simple_loss=0.2907, pruned_loss=0.07042, over 8142.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2977, pruned_loss=0.06978, over 1616767.77 frames. ], batch size: 22, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:38:17,609 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126731.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:38:24,939 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=126742.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:38:28,283 INFO [train.py:901] (1/4) Epoch 16, batch 5500, loss[loss=0.2138, simple_loss=0.3039, pruned_loss=0.06179, over 8450.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2971, pruned_loss=0.06974, over 1614813.95 frames. ], batch size: 29, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:38:28,996 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 18:38:35,365 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126756.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:38:44,224 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.451e+02 2.886e+02 3.496e+02 8.391e+02, threshold=5.772e+02, percent-clipped=4.0 +2023-02-06 18:39:02,249 INFO [train.py:901] (1/4) Epoch 16, batch 5550, loss[loss=0.2786, simple_loss=0.35, pruned_loss=0.1037, over 8244.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2973, pruned_loss=0.06964, over 1617767.36 frames. ], batch size: 22, lr: 4.73e-03, grad_scale: 4.0 +2023-02-06 18:39:13,435 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=126812.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:39:30,332 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126834.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:39:38,324 INFO [train.py:901] (1/4) Epoch 16, batch 5600, loss[loss=0.2225, simple_loss=0.3102, pruned_loss=0.06741, over 8198.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.297, pruned_loss=0.06937, over 1617809.13 frames. ], batch size: 23, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:39:45,809 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126857.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:39:54,359 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.374e+02 2.959e+02 4.088e+02 8.002e+02, threshold=5.917e+02, percent-clipped=4.0 +2023-02-06 18:40:09,610 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1965, 1.9344, 2.6323, 2.1439, 2.5147, 2.1796, 1.8690, 1.1963], + device='cuda:1'), covar=tensor([0.4981, 0.4235, 0.1522, 0.3199, 0.2146, 0.2597, 0.1854, 0.4627], + device='cuda:1'), in_proj_covar=tensor([0.0916, 0.0924, 0.0762, 0.0897, 0.0961, 0.0846, 0.0721, 0.0797], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:40:12,811 INFO [train.py:901] (1/4) Epoch 16, batch 5650, loss[loss=0.227, simple_loss=0.2969, pruned_loss=0.07859, over 7986.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2978, pruned_loss=0.06999, over 1618209.73 frames. ], batch size: 21, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:40:33,384 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 18:40:33,506 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:40:48,555 INFO [train.py:901] (1/4) Epoch 16, batch 5700, loss[loss=0.2154, simple_loss=0.2968, pruned_loss=0.06698, over 8494.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2972, pruned_loss=0.06989, over 1613813.19 frames. ], batch size: 29, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:41:04,180 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.583e+02 3.205e+02 4.543e+02 7.570e+02, threshold=6.410e+02, percent-clipped=11.0 +2023-02-06 18:41:14,333 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8907, 1.5939, 2.0640, 1.7777, 1.9641, 1.8837, 1.6432, 0.7661], + device='cuda:1'), covar=tensor([0.4702, 0.3976, 0.1524, 0.2768, 0.1939, 0.2526, 0.1752, 0.4080], + device='cuda:1'), in_proj_covar=tensor([0.0914, 0.0926, 0.0765, 0.0897, 0.0963, 0.0847, 0.0723, 0.0800], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:41:17,568 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9914, 1.6277, 1.3857, 1.5401, 1.4174, 1.2693, 1.2848, 1.2999], + device='cuda:1'), covar=tensor([0.1177, 0.0502, 0.1260, 0.0554, 0.0776, 0.1527, 0.0927, 0.0795], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0234, 0.0327, 0.0304, 0.0301, 0.0336, 0.0346, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 18:41:22,819 INFO [train.py:901] (1/4) Epoch 16, batch 5750, loss[loss=0.208, simple_loss=0.2967, pruned_loss=0.05967, over 8558.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2949, pruned_loss=0.06849, over 1614362.58 frames. ], batch size: 31, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:41:39,579 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 18:41:56,553 INFO [train.py:901] (1/4) Epoch 16, batch 5800, loss[loss=0.2441, simple_loss=0.314, pruned_loss=0.08715, over 8471.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2949, pruned_loss=0.0683, over 1613779.95 frames. ], batch size: 29, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:42:14,367 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 2.425e+02 2.951e+02 3.537e+02 6.549e+02, threshold=5.902e+02, percent-clipped=1.0 +2023-02-06 18:42:27,503 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 18:42:33,212 INFO [train.py:901] (1/4) Epoch 16, batch 5850, loss[loss=0.174, simple_loss=0.2684, pruned_loss=0.03976, over 8357.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2956, pruned_loss=0.06827, over 1620648.56 frames. ], batch size: 24, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:42:45,185 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127113.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:43:02,066 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127138.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:43:07,120 INFO [train.py:901] (1/4) Epoch 16, batch 5900, loss[loss=0.1875, simple_loss=0.2611, pruned_loss=0.05695, over 8099.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2973, pruned_loss=0.06929, over 1622104.59 frames. ], batch size: 21, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:43:09,478 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2693, 1.9629, 2.6963, 2.2607, 2.7038, 2.2175, 1.9082, 1.3072], + device='cuda:1'), covar=tensor([0.4764, 0.4618, 0.1521, 0.2977, 0.1975, 0.2515, 0.1810, 0.4740], + device='cuda:1'), in_proj_covar=tensor([0.0911, 0.0921, 0.0764, 0.0893, 0.0959, 0.0842, 0.0720, 0.0796], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:43:12,309 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1892, 1.8802, 2.5157, 2.0897, 2.5050, 2.1598, 1.8443, 1.2881], + device='cuda:1'), covar=tensor([0.5325, 0.4915, 0.1771, 0.3516, 0.2490, 0.3061, 0.2195, 0.5027], + device='cuda:1'), in_proj_covar=tensor([0.0911, 0.0922, 0.0764, 0.0894, 0.0960, 0.0843, 0.0721, 0.0796], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:43:22,996 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.337e+02 2.920e+02 3.581e+02 1.365e+03, threshold=5.840e+02, percent-clipped=5.0 +2023-02-06 18:43:30,612 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=127178.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:43:34,125 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127183.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:43:42,723 INFO [train.py:901] (1/4) Epoch 16, batch 5950, loss[loss=0.243, simple_loss=0.3159, pruned_loss=0.08506, over 8097.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2965, pruned_loss=0.06913, over 1617349.73 frames. ], batch size: 23, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:43:51,310 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127208.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:44:17,703 INFO [train.py:901] (1/4) Epoch 16, batch 6000, loss[loss=0.2504, simple_loss=0.316, pruned_loss=0.09243, over 7642.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2975, pruned_loss=0.06982, over 1620886.35 frames. ], batch size: 19, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:44:17,703 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 18:44:29,970 INFO [train.py:935] (1/4) Epoch 16, validation: loss=0.1793, simple_loss=0.2799, pruned_loss=0.03935, over 944034.00 frames. +2023-02-06 18:44:29,972 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 18:44:44,471 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127267.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:44:45,670 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.282e+02 2.976e+02 3.659e+02 8.304e+02, threshold=5.951e+02, percent-clipped=2.0 +2023-02-06 18:44:47,121 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3312, 2.1070, 3.3029, 1.2115, 2.5372, 1.7993, 1.6256, 2.3853], + device='cuda:1'), covar=tensor([0.2181, 0.2541, 0.0842, 0.4779, 0.1689, 0.3472, 0.2294, 0.2436], + device='cuda:1'), in_proj_covar=tensor([0.0502, 0.0554, 0.0538, 0.0609, 0.0622, 0.0562, 0.0498, 0.0616], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 18:45:01,816 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127293.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:45:03,667 INFO [train.py:901] (1/4) Epoch 16, batch 6050, loss[loss=0.26, simple_loss=0.348, pruned_loss=0.08601, over 8462.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.298, pruned_loss=0.06986, over 1620144.76 frames. ], batch size: 25, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:45:39,308 INFO [train.py:901] (1/4) Epoch 16, batch 6100, loss[loss=0.2302, simple_loss=0.3076, pruned_loss=0.07641, over 8123.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2973, pruned_loss=0.06925, over 1617492.21 frames. ], batch size: 22, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:45:55,488 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.555e+02 2.947e+02 3.627e+02 8.036e+02, threshold=5.895e+02, percent-clipped=1.0 +2023-02-06 18:45:59,287 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 18:46:09,096 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 18:46:13,726 INFO [train.py:901] (1/4) Epoch 16, batch 6150, loss[loss=0.2061, simple_loss=0.2877, pruned_loss=0.06224, over 8140.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2972, pruned_loss=0.06924, over 1621340.19 frames. ], batch size: 22, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:46:28,957 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8824, 6.0148, 5.1500, 2.3830, 5.2864, 5.6150, 5.4798, 5.2724], + device='cuda:1'), covar=tensor([0.0477, 0.0356, 0.0966, 0.4343, 0.0761, 0.0820, 0.1108, 0.0579], + device='cuda:1'), in_proj_covar=tensor([0.0500, 0.0413, 0.0415, 0.0518, 0.0407, 0.0415, 0.0407, 0.0358], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 18:46:48,829 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127445.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:46:49,325 INFO [train.py:901] (1/4) Epoch 16, batch 6200, loss[loss=0.1918, simple_loss=0.2638, pruned_loss=0.05984, over 7694.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2966, pruned_loss=0.06894, over 1614666.68 frames. ], batch size: 18, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:46:52,899 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1290, 1.5801, 1.7203, 1.3895, 0.9225, 1.5637, 1.8072, 1.7235], + device='cuda:1'), covar=tensor([0.0459, 0.1150, 0.1669, 0.1367, 0.0594, 0.1430, 0.0654, 0.0566], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0190, 0.0156, 0.0100, 0.0162, 0.0113, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 18:47:04,720 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.656e+02 3.320e+02 4.256e+02 8.643e+02, threshold=6.639e+02, percent-clipped=4.0 +2023-02-06 18:47:23,440 INFO [train.py:901] (1/4) Epoch 16, batch 6250, loss[loss=0.2048, simple_loss=0.2846, pruned_loss=0.06254, over 7714.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2953, pruned_loss=0.06852, over 1612366.45 frames. ], batch size: 18, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:47:57,828 INFO [train.py:901] (1/4) Epoch 16, batch 6300, loss[loss=0.1943, simple_loss=0.2766, pruned_loss=0.05601, over 7425.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2963, pruned_loss=0.06882, over 1616096.98 frames. ], batch size: 17, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:47:58,587 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127547.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:47:59,970 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127549.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:48:08,822 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1763, 1.8704, 2.5776, 2.1057, 2.5390, 2.1778, 1.9022, 1.2256], + device='cuda:1'), covar=tensor([0.4513, 0.4258, 0.1468, 0.3102, 0.2002, 0.2616, 0.1693, 0.4586], + device='cuda:1'), in_proj_covar=tensor([0.0913, 0.0921, 0.0766, 0.0896, 0.0964, 0.0846, 0.0722, 0.0796], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:48:13,472 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2962, 2.6720, 3.0068, 1.5472, 3.1158, 1.9719, 1.4985, 2.1581], + device='cuda:1'), covar=tensor([0.0527, 0.0255, 0.0186, 0.0576, 0.0339, 0.0546, 0.0615, 0.0405], + device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0362, 0.0311, 0.0416, 0.0349, 0.0507, 0.0367, 0.0389], + device='cuda:1'), out_proj_covar=tensor([1.1525e-04, 9.6553e-05, 8.2719e-05, 1.1171e-04, 9.3875e-05, 1.4659e-04, + 1.0041e-04, 1.0499e-04], device='cuda:1') +2023-02-06 18:48:14,536 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.653e+02 3.258e+02 3.936e+02 6.732e+02, threshold=6.516e+02, percent-clipped=2.0 +2023-02-06 18:48:17,991 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127574.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:48:32,719 INFO [train.py:901] (1/4) Epoch 16, batch 6350, loss[loss=0.2025, simple_loss=0.2945, pruned_loss=0.05527, over 8334.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2952, pruned_loss=0.06804, over 1616082.19 frames. ], batch size: 25, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:48:43,690 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=127611.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:49:01,230 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4214, 2.4402, 1.4828, 2.2008, 2.0885, 1.2094, 1.9307, 2.0058], + device='cuda:1'), covar=tensor([0.1614, 0.0545, 0.1490, 0.0655, 0.0821, 0.2134, 0.1183, 0.0952], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0233, 0.0323, 0.0300, 0.0297, 0.0330, 0.0340, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 18:49:03,419 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-06 18:49:07,014 INFO [train.py:901] (1/4) Epoch 16, batch 6400, loss[loss=0.2259, simple_loss=0.3098, pruned_loss=0.07103, over 8341.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2972, pruned_loss=0.06906, over 1621681.92 frames. ], batch size: 26, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:49:18,280 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0444, 3.9125, 2.3405, 2.9072, 3.0321, 2.0012, 3.0328, 3.0956], + device='cuda:1'), covar=tensor([0.1754, 0.0314, 0.1051, 0.0785, 0.0622, 0.1463, 0.0919, 0.0981], + device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0232, 0.0323, 0.0299, 0.0297, 0.0330, 0.0339, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 18:49:24,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.402e+02 3.034e+02 3.710e+02 8.847e+02, threshold=6.069e+02, percent-clipped=1.0 +2023-02-06 18:49:32,558 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127680.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:49:43,222 INFO [train.py:901] (1/4) Epoch 16, batch 6450, loss[loss=0.1922, simple_loss=0.2785, pruned_loss=0.05302, over 8028.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2951, pruned_loss=0.06806, over 1616170.10 frames. ], batch size: 22, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:50:04,137 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127726.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:50:17,017 INFO [train.py:901] (1/4) Epoch 16, batch 6500, loss[loss=0.2649, simple_loss=0.3425, pruned_loss=0.09365, over 8336.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2967, pruned_loss=0.06896, over 1613267.43 frames. ], batch size: 25, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:50:32,626 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.427e+02 3.150e+02 4.006e+02 1.604e+03, threshold=6.301e+02, percent-clipped=4.0 +2023-02-06 18:50:48,406 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=127789.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:50:52,953 INFO [train.py:901] (1/4) Epoch 16, batch 6550, loss[loss=0.1967, simple_loss=0.2745, pruned_loss=0.05948, over 7205.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2955, pruned_loss=0.06835, over 1610702.83 frames. ], batch size: 16, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:50:53,791 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127797.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:51:17,263 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 18:51:27,458 INFO [train.py:901] (1/4) Epoch 16, batch 6600, loss[loss=0.2107, simple_loss=0.291, pruned_loss=0.06519, over 8651.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.295, pruned_loss=0.0681, over 1610669.23 frames. ], batch size: 34, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:51:36,814 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 18:51:42,275 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:51:42,773 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.405e+02 2.899e+02 3.574e+02 1.034e+03, threshold=5.799e+02, percent-clipped=3.0 +2023-02-06 18:51:57,078 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127890.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:51:57,612 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=127891.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:52:00,756 INFO [train.py:901] (1/4) Epoch 16, batch 6650, loss[loss=0.1709, simple_loss=0.2476, pruned_loss=0.04708, over 7651.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2943, pruned_loss=0.0678, over 1609397.95 frames. ], batch size: 19, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:52:07,585 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127904.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:52:36,164 INFO [train.py:901] (1/4) Epoch 16, batch 6700, loss[loss=0.1896, simple_loss=0.2651, pruned_loss=0.05708, over 7709.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2955, pruned_loss=0.06825, over 1611605.08 frames. ], batch size: 18, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:52:41,760 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5091, 1.4847, 1.8376, 1.3311, 1.1495, 1.8153, 0.1633, 1.1849], + device='cuda:1'), covar=tensor([0.1894, 0.1415, 0.0433, 0.1043, 0.3032, 0.0485, 0.2471, 0.1361], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0184, 0.0114, 0.0212, 0.0259, 0.0118, 0.0165, 0.0179], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 18:52:52,430 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.543e+02 2.898e+02 3.564e+02 8.195e+02, threshold=5.796e+02, percent-clipped=3.0 +2023-02-06 18:53:01,214 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127982.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:53:10,440 INFO [train.py:901] (1/4) Epoch 16, batch 6750, loss[loss=0.2077, simple_loss=0.2679, pruned_loss=0.07377, over 7540.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2949, pruned_loss=0.06835, over 1609309.13 frames. ], batch size: 18, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:53:15,090 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0107, 2.0320, 1.8377, 2.5902, 1.1906, 1.5030, 1.8850, 2.0542], + device='cuda:1'), covar=tensor([0.0693, 0.0797, 0.0927, 0.0370, 0.1000, 0.1370, 0.0771, 0.0792], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0201, 0.0247, 0.0211, 0.0208, 0.0247, 0.0252, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 18:53:15,267 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-06 18:53:18,487 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:53:19,174 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128007.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:53:32,878 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128024.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:53:47,639 INFO [train.py:901] (1/4) Epoch 16, batch 6800, loss[loss=0.2002, simple_loss=0.2875, pruned_loss=0.05648, over 8339.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2946, pruned_loss=0.06794, over 1607122.79 frames. ], batch size: 25, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:53:51,073 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 18:54:04,017 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.604e+02 3.143e+02 4.008e+02 8.483e+02, threshold=6.287e+02, percent-clipped=3.0 +2023-02-06 18:54:22,245 INFO [train.py:901] (1/4) Epoch 16, batch 6850, loss[loss=0.2437, simple_loss=0.3236, pruned_loss=0.08185, over 8103.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2954, pruned_loss=0.06835, over 1607890.28 frames. ], batch size: 23, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:54:40,675 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 18:54:53,394 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128139.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:54:54,709 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128141.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:54:56,196 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6930, 1.8294, 2.6019, 1.4968, 1.1733, 2.4580, 0.3377, 1.4585], + device='cuda:1'), covar=tensor([0.2151, 0.1540, 0.0438, 0.2194, 0.3588, 0.0468, 0.2784, 0.1911], + device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0183, 0.0114, 0.0212, 0.0258, 0.0118, 0.0164, 0.0178], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 18:54:58,084 INFO [train.py:901] (1/4) Epoch 16, batch 6900, loss[loss=0.2428, simple_loss=0.3306, pruned_loss=0.07752, over 8025.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2944, pruned_loss=0.06755, over 1608487.93 frames. ], batch size: 22, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:55:08,137 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128160.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:55:14,249 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.605e+02 3.172e+02 3.868e+02 9.306e+02, threshold=6.344e+02, percent-clipped=5.0 +2023-02-06 18:55:25,857 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128185.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:55:32,787 INFO [train.py:901] (1/4) Epoch 16, batch 6950, loss[loss=0.2582, simple_loss=0.3312, pruned_loss=0.09257, over 8587.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2948, pruned_loss=0.0675, over 1612053.58 frames. ], batch size: 34, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:55:34,511 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 18:55:43,518 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128212.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:55:48,029 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 18:55:58,348 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128234.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:56:07,210 INFO [train.py:901] (1/4) Epoch 16, batch 7000, loss[loss=0.1843, simple_loss=0.2561, pruned_loss=0.05621, over 7260.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.295, pruned_loss=0.06769, over 1615458.70 frames. ], batch size: 16, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:56:15,595 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128256.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:56:19,645 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128262.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:56:24,008 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.690e+02 3.457e+02 5.056e+02 8.270e+02, threshold=6.915e+02, percent-clipped=6.0 +2023-02-06 18:56:36,190 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128287.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:56:42,585 INFO [train.py:901] (1/4) Epoch 16, batch 7050, loss[loss=0.2549, simple_loss=0.3234, pruned_loss=0.09318, over 8240.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2961, pruned_loss=0.06862, over 1614204.93 frames. ], batch size: 22, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:57:03,871 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:57:11,329 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:57:16,468 INFO [train.py:901] (1/4) Epoch 16, batch 7100, loss[loss=0.1837, simple_loss=0.262, pruned_loss=0.05272, over 7813.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2963, pruned_loss=0.06863, over 1612417.36 frames. ], batch size: 20, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:57:18,653 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:57:33,888 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 2.456e+02 3.083e+02 3.766e+02 8.441e+02, threshold=6.166e+02, percent-clipped=2.0 +2023-02-06 18:57:52,032 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128395.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:57:52,512 INFO [train.py:901] (1/4) Epoch 16, batch 7150, loss[loss=0.2492, simple_loss=0.3193, pruned_loss=0.0896, over 8478.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2954, pruned_loss=0.06776, over 1610615.30 frames. ], batch size: 49, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:58:08,447 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1969, 2.2083, 1.7255, 1.9277, 1.7751, 1.3948, 1.6112, 1.6757], + device='cuda:1'), covar=tensor([0.1205, 0.0343, 0.1033, 0.0558, 0.0731, 0.1427, 0.0967, 0.0811], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0233, 0.0325, 0.0301, 0.0302, 0.0330, 0.0342, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 18:58:09,831 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128420.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:58:17,229 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128431.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:58:27,263 INFO [train.py:901] (1/4) Epoch 16, batch 7200, loss[loss=0.2575, simple_loss=0.3307, pruned_loss=0.09211, over 8622.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2956, pruned_loss=0.06829, over 1603577.83 frames. ], batch size: 50, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:58:42,533 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.378e+02 2.905e+02 3.370e+02 6.119e+02, threshold=5.810e+02, percent-clipped=0.0 +2023-02-06 18:59:02,793 INFO [train.py:901] (1/4) Epoch 16, batch 7250, loss[loss=0.2331, simple_loss=0.3096, pruned_loss=0.07826, over 8288.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2959, pruned_loss=0.06849, over 1606075.88 frames. ], batch size: 23, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:59:13,764 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128512.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:59:26,460 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1735, 1.3359, 4.3455, 1.7075, 3.8359, 3.5874, 3.9428, 3.7910], + device='cuda:1'), covar=tensor([0.0466, 0.4703, 0.0518, 0.3634, 0.1116, 0.0937, 0.0486, 0.0631], + device='cuda:1'), in_proj_covar=tensor([0.0561, 0.0623, 0.0643, 0.0591, 0.0670, 0.0573, 0.0564, 0.0632], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 18:59:31,352 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128537.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:59:32,910 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 18:59:37,230 INFO [train.py:901] (1/4) Epoch 16, batch 7300, loss[loss=0.2, simple_loss=0.2877, pruned_loss=0.05616, over 8562.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2955, pruned_loss=0.06809, over 1607363.29 frames. ], batch size: 31, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:59:52,609 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.470e+02 2.980e+02 3.722e+02 1.252e+03, threshold=5.960e+02, percent-clipped=4.0 +2023-02-06 19:00:02,295 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:00:12,482 INFO [train.py:901] (1/4) Epoch 16, batch 7350, loss[loss=0.2273, simple_loss=0.3063, pruned_loss=0.07417, over 8358.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2954, pruned_loss=0.06827, over 1608027.73 frames. ], batch size: 24, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 19:00:19,507 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:00:21,498 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:00:31,402 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 19:00:36,122 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:00:47,435 INFO [train.py:901] (1/4) Epoch 16, batch 7400, loss[loss=0.1968, simple_loss=0.2746, pruned_loss=0.05949, over 7556.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2963, pruned_loss=0.06864, over 1607873.52 frames. ], batch size: 18, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 19:00:49,533 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 19:01:00,966 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5141, 1.3752, 1.7201, 1.3102, 0.9380, 1.5594, 1.5568, 1.4324], + device='cuda:1'), covar=tensor([0.0533, 0.1240, 0.1640, 0.1388, 0.0595, 0.1455, 0.0672, 0.0632], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0189, 0.0156, 0.0100, 0.0161, 0.0114, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 19:01:02,829 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.610e+02 3.305e+02 3.788e+02 1.058e+03, threshold=6.610e+02, percent-clipped=7.0 +2023-02-06 19:01:11,760 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128682.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:01:21,080 INFO [train.py:901] (1/4) Epoch 16, batch 7450, loss[loss=0.2281, simple_loss=0.3099, pruned_loss=0.07316, over 8643.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2957, pruned_loss=0.06845, over 1609573.95 frames. ], batch size: 34, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 19:01:30,639 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 19:01:52,242 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8919, 6.0701, 5.2458, 2.2978, 5.3215, 5.7662, 5.5147, 5.4532], + device='cuda:1'), covar=tensor([0.0737, 0.0433, 0.1038, 0.5368, 0.0747, 0.0722, 0.1237, 0.0594], + device='cuda:1'), in_proj_covar=tensor([0.0503, 0.0411, 0.0416, 0.0516, 0.0403, 0.0412, 0.0405, 0.0357], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:01:56,805 INFO [train.py:901] (1/4) Epoch 16, batch 7500, loss[loss=0.2122, simple_loss=0.2972, pruned_loss=0.06359, over 8489.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.295, pruned_loss=0.06803, over 1610134.48 frames. ], batch size: 29, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 19:02:12,296 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-06 19:02:13,137 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 2.417e+02 2.923e+02 3.614e+02 6.549e+02, threshold=5.847e+02, percent-clipped=0.0 +2023-02-06 19:02:17,186 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128775.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:02:31,137 INFO [train.py:901] (1/4) Epoch 16, batch 7550, loss[loss=0.2254, simple_loss=0.2871, pruned_loss=0.08185, over 6833.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2941, pruned_loss=0.06753, over 1608482.33 frames. ], batch size: 15, lr: 4.69e-03, grad_scale: 16.0 +2023-02-06 19:02:32,034 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128797.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:02:33,331 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128799.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:02:49,736 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5814, 1.7576, 2.6662, 1.3650, 1.9786, 1.9238, 1.5414, 1.9017], + device='cuda:1'), covar=tensor([0.1768, 0.2459, 0.0959, 0.4215, 0.1761, 0.2940, 0.2174, 0.2107], + device='cuda:1'), in_proj_covar=tensor([0.0507, 0.0562, 0.0546, 0.0613, 0.0630, 0.0570, 0.0504, 0.0620], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:03:04,845 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2045, 1.1061, 1.2817, 1.0407, 0.9842, 1.3046, 0.0459, 0.8809], + device='cuda:1'), covar=tensor([0.2126, 0.1590, 0.0531, 0.0992, 0.3132, 0.0626, 0.2535, 0.1406], + device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0185, 0.0115, 0.0215, 0.0261, 0.0120, 0.0165, 0.0179], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 19:03:07,389 INFO [train.py:901] (1/4) Epoch 16, batch 7600, loss[loss=0.1965, simple_loss=0.2779, pruned_loss=0.05753, over 8185.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2943, pruned_loss=0.06758, over 1607862.81 frames. ], batch size: 23, lr: 4.69e-03, grad_scale: 16.0 +2023-02-06 19:03:23,266 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 2.439e+02 3.123e+02 4.017e+02 8.994e+02, threshold=6.245e+02, percent-clipped=5.0 +2023-02-06 19:03:38,585 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128890.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:03:42,461 INFO [train.py:901] (1/4) Epoch 16, batch 7650, loss[loss=0.1994, simple_loss=0.2862, pruned_loss=0.05631, over 8115.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2949, pruned_loss=0.06767, over 1610763.26 frames. ], batch size: 23, lr: 4.69e-03, grad_scale: 16.0 +2023-02-06 19:03:50,523 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128908.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:04:08,860 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9381, 6.1241, 5.3248, 2.3723, 5.3571, 5.7358, 5.4823, 5.3904], + device='cuda:1'), covar=tensor([0.0588, 0.0401, 0.1050, 0.4381, 0.0716, 0.0733, 0.1202, 0.0394], + device='cuda:1'), in_proj_covar=tensor([0.0497, 0.0406, 0.0411, 0.0508, 0.0400, 0.0408, 0.0400, 0.0353], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:04:17,495 INFO [train.py:901] (1/4) Epoch 16, batch 7700, loss[loss=0.205, simple_loss=0.2819, pruned_loss=0.06409, over 8236.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2955, pruned_loss=0.06832, over 1612415.82 frames. ], batch size: 22, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:04:34,542 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.361e+02 3.016e+02 3.880e+02 7.767e+02, threshold=6.032e+02, percent-clipped=3.0 +2023-02-06 19:04:42,144 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 19:04:49,725 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4944, 1.9253, 3.2273, 1.3162, 2.5837, 1.8867, 1.6388, 2.3354], + device='cuda:1'), covar=tensor([0.1913, 0.2502, 0.0869, 0.4312, 0.1622, 0.3028, 0.2160, 0.2283], + device='cuda:1'), in_proj_covar=tensor([0.0504, 0.0561, 0.0544, 0.0612, 0.0628, 0.0570, 0.0502, 0.0617], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:04:52,877 INFO [train.py:901] (1/4) Epoch 16, batch 7750, loss[loss=0.1769, simple_loss=0.2479, pruned_loss=0.05296, over 7705.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2946, pruned_loss=0.06749, over 1611984.55 frames. ], batch size: 18, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:04:56,379 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6571, 1.8218, 2.7083, 1.4719, 1.9941, 1.9703, 1.7527, 1.7493], + device='cuda:1'), covar=tensor([0.1763, 0.2665, 0.0858, 0.4398, 0.1788, 0.3044, 0.2116, 0.2311], + device='cuda:1'), in_proj_covar=tensor([0.0505, 0.0562, 0.0545, 0.0613, 0.0629, 0.0571, 0.0503, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:05:21,214 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5915, 1.6624, 1.7295, 1.3301, 1.8082, 1.4281, 0.9153, 1.5823], + device='cuda:1'), covar=tensor([0.0426, 0.0264, 0.0181, 0.0391, 0.0305, 0.0572, 0.0633, 0.0241], + device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0364, 0.0314, 0.0421, 0.0350, 0.0510, 0.0373, 0.0393], + device='cuda:1'), out_proj_covar=tensor([1.1664e-04, 9.6988e-05, 8.3333e-05, 1.1296e-04, 9.4072e-05, 1.4731e-04, + 1.0217e-04, 1.0608e-04], device='cuda:1') +2023-02-06 19:05:24,791 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 19:05:26,241 INFO [train.py:901] (1/4) Epoch 16, batch 7800, loss[loss=0.2419, simple_loss=0.3188, pruned_loss=0.08253, over 7971.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2965, pruned_loss=0.06886, over 1610010.88 frames. ], batch size: 21, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:05:28,997 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1465, 2.1399, 1.5764, 1.8916, 1.8200, 1.2790, 1.5006, 1.7040], + device='cuda:1'), covar=tensor([0.1330, 0.0353, 0.1135, 0.0558, 0.0720, 0.1554, 0.0982, 0.0874], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0234, 0.0327, 0.0300, 0.0299, 0.0332, 0.0341, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 19:05:31,087 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129053.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:05:41,944 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.423e+02 2.949e+02 3.975e+02 9.373e+02, threshold=5.898e+02, percent-clipped=5.0 +2023-02-06 19:05:45,474 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3029, 1.2686, 2.2816, 1.0941, 2.1141, 2.4749, 2.5859, 1.9392], + device='cuda:1'), covar=tensor([0.1172, 0.1456, 0.0569, 0.2360, 0.0852, 0.0451, 0.0786, 0.0974], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0309, 0.0273, 0.0301, 0.0293, 0.0251, 0.0385, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 19:05:48,209 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129078.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:06:00,669 INFO [train.py:901] (1/4) Epoch 16, batch 7850, loss[loss=0.2167, simple_loss=0.2916, pruned_loss=0.07094, over 7966.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2976, pruned_loss=0.06971, over 1614184.62 frames. ], batch size: 21, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:06:09,643 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8790, 3.8116, 3.4419, 1.6740, 3.3740, 3.4183, 3.5245, 3.1762], + device='cuda:1'), covar=tensor([0.0844, 0.0604, 0.1169, 0.4589, 0.0932, 0.1161, 0.1261, 0.0958], + device='cuda:1'), in_proj_covar=tensor([0.0505, 0.0413, 0.0420, 0.0516, 0.0407, 0.0415, 0.0407, 0.0360], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:06:32,323 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=129143.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:06:34,271 INFO [train.py:901] (1/4) Epoch 16, batch 7900, loss[loss=0.1925, simple_loss=0.2625, pruned_loss=0.06121, over 6788.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2969, pruned_loss=0.06908, over 1611482.10 frames. ], batch size: 15, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:06:34,504 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129146.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:06:51,032 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.289e+02 2.786e+02 3.620e+02 6.776e+02, threshold=5.572e+02, percent-clipped=2.0 +2023-02-06 19:06:51,878 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129171.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:07:08,441 INFO [train.py:901] (1/4) Epoch 16, batch 7950, loss[loss=0.232, simple_loss=0.3122, pruned_loss=0.07588, over 8421.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2958, pruned_loss=0.06835, over 1609779.50 frames. ], batch size: 29, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:07:42,911 INFO [train.py:901] (1/4) Epoch 16, batch 8000, loss[loss=0.1917, simple_loss=0.2687, pruned_loss=0.05737, over 7420.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2951, pruned_loss=0.06758, over 1613149.23 frames. ], batch size: 17, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:07:47,178 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=129252.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:07:48,166 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.31 vs. limit=5.0 +2023-02-06 19:07:51,187 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129258.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:07:59,068 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.511e+02 2.964e+02 3.601e+02 8.820e+02, threshold=5.927e+02, percent-clipped=6.0 +2023-02-06 19:08:16,585 INFO [train.py:901] (1/4) Epoch 16, batch 8050, loss[loss=0.284, simple_loss=0.3404, pruned_loss=0.1139, over 6659.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.294, pruned_loss=0.0684, over 1595449.24 frames. ], batch size: 72, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:08:53,103 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 19:08:56,566 INFO [train.py:901] (1/4) Epoch 17, batch 0, loss[loss=0.2459, simple_loss=0.3319, pruned_loss=0.07995, over 8187.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3319, pruned_loss=0.07995, over 8187.00 frames. ], batch size: 23, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:08:56,566 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 19:09:04,778 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5326, 1.7732, 2.6447, 1.3627, 1.9236, 1.8299, 1.6015, 1.8623], + device='cuda:1'), covar=tensor([0.1739, 0.2474, 0.0884, 0.4321, 0.1796, 0.3066, 0.2173, 0.2181], + device='cuda:1'), in_proj_covar=tensor([0.0504, 0.0561, 0.0543, 0.0612, 0.0631, 0.0570, 0.0501, 0.0620], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:09:07,566 INFO [train.py:935] (1/4) Epoch 17, validation: loss=0.1792, simple_loss=0.2794, pruned_loss=0.03944, over 944034.00 frames. +2023-02-06 19:09:07,567 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 19:09:19,460 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 19:09:21,132 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 19:09:29,754 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3436, 1.9656, 2.5991, 2.1121, 2.4055, 2.2874, 2.0033, 1.2242], + device='cuda:1'), covar=tensor([0.4235, 0.4200, 0.1585, 0.3107, 0.2219, 0.2607, 0.1722, 0.4936], + device='cuda:1'), in_proj_covar=tensor([0.0906, 0.0918, 0.0758, 0.0892, 0.0954, 0.0845, 0.0719, 0.0792], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 19:09:33,854 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129367.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:09:35,632 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.554e+02 2.551e+02 3.127e+02 3.678e+02 8.568e+02, threshold=6.254e+02, percent-clipped=4.0 +2023-02-06 19:09:41,817 INFO [train.py:901] (1/4) Epoch 17, batch 50, loss[loss=0.2156, simple_loss=0.3028, pruned_loss=0.06419, over 8249.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2985, pruned_loss=0.07018, over 365293.48 frames. ], batch size: 24, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:09:54,009 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 19:10:18,428 INFO [train.py:901] (1/4) Epoch 17, batch 100, loss[loss=0.2154, simple_loss=0.298, pruned_loss=0.06639, over 8461.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2985, pruned_loss=0.07026, over 641695.60 frames. ], batch size: 25, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:10:18,446 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 19:10:19,955 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129431.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:10:32,096 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7764, 3.2261, 2.0513, 2.4554, 2.3557, 1.6589, 2.4086, 2.7312], + device='cuda:1'), covar=tensor([0.1750, 0.0405, 0.1263, 0.0795, 0.0841, 0.1724, 0.1121, 0.1035], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0235, 0.0328, 0.0301, 0.0299, 0.0335, 0.0342, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 19:10:44,164 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6644, 1.9220, 2.0179, 1.3073, 2.2164, 1.4452, 0.6674, 1.7828], + device='cuda:1'), covar=tensor([0.0552, 0.0317, 0.0269, 0.0493, 0.0358, 0.0776, 0.0803, 0.0304], + device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0365, 0.0315, 0.0422, 0.0348, 0.0513, 0.0374, 0.0390], + device='cuda:1'), out_proj_covar=tensor([1.1630e-04, 9.7486e-05, 8.3561e-05, 1.1323e-04, 9.3624e-05, 1.4825e-04, + 1.0240e-04, 1.0518e-04], device='cuda:1') +2023-02-06 19:10:46,022 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.389e+02 2.489e+02 3.062e+02 3.657e+02 7.822e+02, threshold=6.124e+02, percent-clipped=4.0 +2023-02-06 19:10:52,181 INFO [train.py:901] (1/4) Epoch 17, batch 150, loss[loss=0.244, simple_loss=0.3162, pruned_loss=0.08589, over 8323.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2972, pruned_loss=0.06895, over 857518.66 frames. ], batch size: 25, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:11:18,275 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129514.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:11:29,032 INFO [train.py:901] (1/4) Epoch 17, batch 200, loss[loss=0.2502, simple_loss=0.3343, pruned_loss=0.08304, over 8617.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2987, pruned_loss=0.07003, over 1022811.86 frames. ], batch size: 31, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:11:36,219 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129539.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:11:57,069 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.455e+02 2.902e+02 3.926e+02 7.649e+02, threshold=5.804e+02, percent-clipped=5.0 +2023-02-06 19:12:03,425 INFO [train.py:901] (1/4) Epoch 17, batch 250, loss[loss=0.218, simple_loss=0.3065, pruned_loss=0.06476, over 8499.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2977, pruned_loss=0.06861, over 1158740.98 frames. ], batch size: 28, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:12:09,649 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 19:12:11,838 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9312, 1.5990, 1.3488, 1.5490, 1.2725, 1.2181, 1.1816, 1.2823], + device='cuda:1'), covar=tensor([0.1136, 0.0408, 0.1212, 0.0493, 0.0721, 0.1480, 0.0917, 0.0718], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0236, 0.0330, 0.0303, 0.0301, 0.0337, 0.0344, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 19:12:18,387 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 19:12:33,727 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129623.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:12:38,235 INFO [train.py:901] (1/4) Epoch 17, batch 300, loss[loss=0.2087, simple_loss=0.3024, pruned_loss=0.05748, over 8298.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2982, pruned_loss=0.06981, over 1255016.98 frames. ], batch size: 23, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:12:39,081 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:12:53,545 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129648.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:13:08,138 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.453e+02 3.064e+02 3.747e+02 1.027e+03, threshold=6.129e+02, percent-clipped=5.0 +2023-02-06 19:13:14,335 INFO [train.py:901] (1/4) Epoch 17, batch 350, loss[loss=0.2622, simple_loss=0.3424, pruned_loss=0.09105, over 8603.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2969, pruned_loss=0.06912, over 1329793.56 frames. ], batch size: 34, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:13:47,829 INFO [train.py:901] (1/4) Epoch 17, batch 400, loss[loss=0.1997, simple_loss=0.2709, pruned_loss=0.06419, over 7661.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2956, pruned_loss=0.06854, over 1396312.95 frames. ], batch size: 19, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:14:08,512 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0356, 1.5219, 1.7677, 1.4560, 0.9509, 1.5624, 1.8088, 1.6651], + device='cuda:1'), covar=tensor([0.0493, 0.1210, 0.1624, 0.1354, 0.0586, 0.1423, 0.0670, 0.0606], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0189, 0.0155, 0.0100, 0.0161, 0.0114, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 19:14:18,000 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.355e+02 2.898e+02 3.830e+02 8.224e+02, threshold=5.797e+02, percent-clipped=7.0 +2023-02-06 19:14:21,452 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=129775.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:14:24,069 INFO [train.py:901] (1/4) Epoch 17, batch 450, loss[loss=0.1908, simple_loss=0.2622, pruned_loss=0.05973, over 7518.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2953, pruned_loss=0.06843, over 1444315.25 frames. ], batch size: 18, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:14:26,519 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-06 19:14:39,896 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.07 vs. limit=5.0 +2023-02-06 19:14:44,133 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 19:14:44,497 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129809.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:14:58,034 INFO [train.py:901] (1/4) Epoch 17, batch 500, loss[loss=0.2112, simple_loss=0.2909, pruned_loss=0.0657, over 8138.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2964, pruned_loss=0.06911, over 1485708.32 frames. ], batch size: 22, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:15:28,004 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.377e+02 2.910e+02 3.862e+02 1.132e+03, threshold=5.820e+02, percent-clipped=8.0 +2023-02-06 19:15:35,664 INFO [train.py:901] (1/4) Epoch 17, batch 550, loss[loss=0.2378, simple_loss=0.3142, pruned_loss=0.08066, over 8108.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2966, pruned_loss=0.06835, over 1516411.33 frames. ], batch size: 23, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:15:43,349 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129890.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:16:10,106 INFO [train.py:901] (1/4) Epoch 17, batch 600, loss[loss=0.2147, simple_loss=0.2838, pruned_loss=0.07283, over 7437.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2961, pruned_loss=0.06839, over 1536145.71 frames. ], batch size: 17, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:16:19,710 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 19:16:24,102 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5640, 1.3915, 1.5486, 1.3007, 0.9192, 1.3426, 1.5240, 1.3568], + device='cuda:1'), covar=tensor([0.0549, 0.1258, 0.1695, 0.1400, 0.0575, 0.1516, 0.0699, 0.0655], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0190, 0.0156, 0.0100, 0.0162, 0.0114, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 19:16:38,513 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.845e+02 2.576e+02 2.936e+02 3.639e+02 7.352e+02, threshold=5.872e+02, percent-clipped=2.0 +2023-02-06 19:16:41,357 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=129974.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:16:44,765 INFO [train.py:901] (1/4) Epoch 17, batch 650, loss[loss=0.2029, simple_loss=0.2884, pruned_loss=0.0587, over 8461.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2957, pruned_loss=0.06856, over 1556290.96 frames. ], batch size: 25, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:17:09,803 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0888, 2.5893, 3.6738, 1.9250, 1.9220, 3.5349, 0.7450, 2.1957], + device='cuda:1'), covar=tensor([0.1504, 0.1297, 0.0266, 0.2023, 0.3023, 0.0373, 0.2446, 0.1391], + device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0185, 0.0115, 0.0217, 0.0265, 0.0121, 0.0165, 0.0179], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 19:17:11,722 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4886, 1.3749, 4.3151, 1.9607, 2.4905, 4.8575, 4.8485, 4.1942], + device='cuda:1'), covar=tensor([0.0963, 0.1881, 0.0283, 0.2004, 0.1112, 0.0195, 0.0589, 0.0543], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0306, 0.0270, 0.0298, 0.0291, 0.0251, 0.0384, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 19:17:16,419 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:17:16,517 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7962, 2.3676, 4.3878, 1.5926, 3.2896, 2.3452, 1.9252, 3.0867], + device='cuda:1'), covar=tensor([0.1702, 0.2290, 0.0705, 0.3995, 0.1412, 0.2773, 0.1880, 0.2129], + device='cuda:1'), in_proj_covar=tensor([0.0505, 0.0560, 0.0542, 0.0610, 0.0628, 0.0568, 0.0502, 0.0617], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:17:17,843 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130020.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:17:23,794 INFO [train.py:901] (1/4) Epoch 17, batch 700, loss[loss=0.2351, simple_loss=0.3139, pruned_loss=0.07819, over 8481.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2953, pruned_loss=0.06859, over 1569221.19 frames. ], batch size: 25, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:17:51,873 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.350e+02 2.811e+02 3.683e+02 1.098e+03, threshold=5.622e+02, percent-clipped=6.0 +2023-02-06 19:17:58,276 INFO [train.py:901] (1/4) Epoch 17, batch 750, loss[loss=0.1579, simple_loss=0.2528, pruned_loss=0.03153, over 7812.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2963, pruned_loss=0.0685, over 1581140.80 frames. ], batch size: 20, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:18:04,872 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7090, 1.9911, 2.0933, 1.3894, 2.2350, 1.5726, 0.7433, 1.8906], + device='cuda:1'), covar=tensor([0.0444, 0.0261, 0.0208, 0.0429, 0.0295, 0.0663, 0.0685, 0.0238], + device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0362, 0.0309, 0.0418, 0.0346, 0.0507, 0.0370, 0.0388], + device='cuda:1'), out_proj_covar=tensor([1.1516e-04, 9.6632e-05, 8.1938e-05, 1.1221e-04, 9.3132e-05, 1.4651e-04, + 1.0125e-04, 1.0452e-04], device='cuda:1') +2023-02-06 19:18:05,543 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130089.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:18:08,227 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 19:18:19,441 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 19:18:36,021 INFO [train.py:901] (1/4) Epoch 17, batch 800, loss[loss=0.2404, simple_loss=0.3198, pruned_loss=0.08054, over 8336.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2966, pruned_loss=0.06875, over 1590553.25 frames. ], batch size: 25, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:18:46,641 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7636, 1.8548, 2.4385, 1.7330, 1.3934, 2.3965, 0.4227, 1.4543], + device='cuda:1'), covar=tensor([0.2154, 0.1216, 0.0396, 0.1681, 0.3157, 0.0484, 0.2581, 0.1869], + device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0186, 0.0116, 0.0218, 0.0265, 0.0122, 0.0167, 0.0181], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 19:18:48,061 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130146.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:18:52,756 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:19:04,215 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.363e+02 2.676e+02 3.408e+02 8.560e+02, threshold=5.353e+02, percent-clipped=3.0 +2023-02-06 19:19:05,147 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130171.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:19:10,476 INFO [train.py:901] (1/4) Epoch 17, batch 850, loss[loss=0.2214, simple_loss=0.3059, pruned_loss=0.06845, over 8829.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2966, pruned_loss=0.06887, over 1598139.08 frames. ], batch size: 40, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:19:47,568 INFO [train.py:901] (1/4) Epoch 17, batch 900, loss[loss=0.169, simple_loss=0.2404, pruned_loss=0.04882, over 7407.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2968, pruned_loss=0.06871, over 1599660.47 frames. ], batch size: 17, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:20:15,370 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:20:16,500 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.489e+02 3.023e+02 3.878e+02 8.176e+02, threshold=6.045e+02, percent-clipped=7.0 +2023-02-06 19:20:22,807 INFO [train.py:901] (1/4) Epoch 17, batch 950, loss[loss=0.1883, simple_loss=0.2694, pruned_loss=0.05355, over 7260.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2965, pruned_loss=0.06882, over 1601326.09 frames. ], batch size: 16, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:20:29,182 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6267, 2.0793, 3.4998, 1.4116, 2.6465, 2.1485, 1.6470, 2.5414], + device='cuda:1'), covar=tensor([0.1820, 0.2438, 0.0736, 0.4390, 0.1567, 0.2842, 0.2114, 0.2140], + device='cuda:1'), in_proj_covar=tensor([0.0507, 0.0564, 0.0545, 0.0612, 0.0632, 0.0570, 0.0504, 0.0619], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:20:43,381 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 19:20:57,172 INFO [train.py:901] (1/4) Epoch 17, batch 1000, loss[loss=0.1757, simple_loss=0.2447, pruned_loss=0.05339, over 7420.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2966, pruned_loss=0.06881, over 1604953.03 frames. ], batch size: 17, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:21:04,859 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0498, 2.2964, 1.8162, 2.9518, 1.5113, 1.6468, 1.9934, 2.3471], + device='cuda:1'), covar=tensor([0.0725, 0.0756, 0.1075, 0.0344, 0.1122, 0.1418, 0.0981, 0.0803], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0202, 0.0252, 0.0213, 0.0210, 0.0250, 0.0255, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 19:21:09,225 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130345.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:21:20,030 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 19:21:21,949 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130362.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:21:23,312 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:21:27,492 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.680e+02 3.059e+02 3.924e+02 8.380e+02, threshold=6.118e+02, percent-clipped=2.0 +2023-02-06 19:21:27,758 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130370.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:21:33,150 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 19:21:33,589 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 19:21:33,835 INFO [train.py:901] (1/4) Epoch 17, batch 1050, loss[loss=0.1799, simple_loss=0.2613, pruned_loss=0.04922, over 7697.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.296, pruned_loss=0.06842, over 1604441.89 frames. ], batch size: 18, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:21:49,933 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130402.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:08,459 INFO [train.py:901] (1/4) Epoch 17, batch 1100, loss[loss=0.2382, simple_loss=0.3145, pruned_loss=0.08095, over 7969.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2967, pruned_loss=0.06909, over 1606855.47 frames. ], batch size: 21, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:22:14,699 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0882, 1.6826, 3.4274, 1.4106, 2.4614, 3.7906, 3.8338, 3.2176], + device='cuda:1'), covar=tensor([0.1012, 0.1603, 0.0352, 0.2070, 0.0962, 0.0222, 0.0543, 0.0564], + device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0307, 0.0272, 0.0301, 0.0293, 0.0252, 0.0386, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 19:22:23,046 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130450.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:27,196 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130456.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:38,663 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.545e+02 2.978e+02 3.676e+02 6.168e+02, threshold=5.956e+02, percent-clipped=1.0 +2023-02-06 19:22:44,133 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:45,339 INFO [train.py:901] (1/4) Epoch 17, batch 1150, loss[loss=0.2794, simple_loss=0.343, pruned_loss=0.1079, over 6500.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2958, pruned_loss=0.06909, over 1604405.14 frames. ], batch size: 71, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:22:45,510 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:45,990 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 19:23:16,164 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:23:19,418 INFO [train.py:901] (1/4) Epoch 17, batch 1200, loss[loss=0.2376, simple_loss=0.3134, pruned_loss=0.08092, over 8508.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2948, pruned_loss=0.06832, over 1607718.98 frames. ], batch size: 26, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:23:33,411 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130549.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:23:45,144 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130566.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:23:47,781 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.314e+02 2.862e+02 3.617e+02 1.013e+03, threshold=5.724e+02, percent-clipped=2.0 +2023-02-06 19:23:53,882 INFO [train.py:901] (1/4) Epoch 17, batch 1250, loss[loss=0.2278, simple_loss=0.3163, pruned_loss=0.06967, over 7972.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2946, pruned_loss=0.06761, over 1611569.20 frames. ], batch size: 21, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:23:57,433 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:24:07,971 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7598, 1.9778, 2.1913, 1.3307, 2.2929, 1.4252, 0.7159, 1.9625], + device='cuda:1'), covar=tensor([0.0615, 0.0307, 0.0228, 0.0525, 0.0363, 0.0829, 0.0770, 0.0303], + device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0365, 0.0312, 0.0424, 0.0350, 0.0516, 0.0377, 0.0394], + device='cuda:1'), out_proj_covar=tensor([1.1675e-04, 9.7352e-05, 8.2703e-05, 1.1359e-04, 9.3933e-05, 1.4911e-04, + 1.0301e-04, 1.0598e-04], device='cuda:1') +2023-02-06 19:24:30,844 INFO [train.py:901] (1/4) Epoch 17, batch 1300, loss[loss=0.2583, simple_loss=0.3344, pruned_loss=0.09113, over 8315.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2963, pruned_loss=0.0688, over 1615182.66 frames. ], batch size: 25, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:24:35,121 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4323, 2.6709, 3.0556, 1.6008, 3.2573, 1.8737, 1.6024, 2.3383], + device='cuda:1'), covar=tensor([0.0644, 0.0345, 0.0218, 0.0677, 0.0390, 0.0832, 0.0793, 0.0456], + device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0365, 0.0313, 0.0424, 0.0348, 0.0516, 0.0376, 0.0393], + device='cuda:1'), out_proj_covar=tensor([1.1640e-04, 9.7338e-05, 8.2955e-05, 1.1361e-04, 9.3577e-05, 1.4901e-04, + 1.0280e-04, 1.0569e-04], device='cuda:1') +2023-02-06 19:24:59,331 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.380e+02 3.126e+02 3.675e+02 7.509e+02, threshold=6.253e+02, percent-clipped=2.0 +2023-02-06 19:25:05,695 INFO [train.py:901] (1/4) Epoch 17, batch 1350, loss[loss=0.2134, simple_loss=0.2918, pruned_loss=0.06746, over 7924.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2964, pruned_loss=0.06878, over 1614806.11 frames. ], batch size: 20, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:25:06,300 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-06 19:25:43,059 INFO [train.py:901] (1/4) Epoch 17, batch 1400, loss[loss=0.1962, simple_loss=0.2724, pruned_loss=0.05997, over 7929.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2973, pruned_loss=0.06939, over 1615013.88 frames. ], batch size: 20, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:25:46,006 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:25:47,358 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130735.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:25:54,826 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130746.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:03,092 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130758.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:04,424 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:11,013 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 2.607e+02 3.260e+02 4.191e+02 1.113e+03, threshold=6.520e+02, percent-clipped=3.0 +2023-02-06 19:26:17,366 INFO [train.py:901] (1/4) Epoch 17, batch 1450, loss[loss=0.2024, simple_loss=0.2954, pruned_loss=0.05467, over 8460.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2967, pruned_loss=0.06893, over 1616183.61 frames. ], batch size: 29, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:26:20,729 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 19:26:27,819 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:32,210 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130800.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:34,989 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1450, 1.4886, 4.3397, 1.7002, 3.8937, 3.6137, 3.9970, 3.8310], + device='cuda:1'), covar=tensor([0.0570, 0.4343, 0.0500, 0.3488, 0.1054, 0.0929, 0.0517, 0.0666], + device='cuda:1'), in_proj_covar=tensor([0.0570, 0.0627, 0.0650, 0.0595, 0.0676, 0.0580, 0.0573, 0.0637], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 19:26:54,254 INFO [train.py:901] (1/4) Epoch 17, batch 1500, loss[loss=0.1974, simple_loss=0.2784, pruned_loss=0.0582, over 8089.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2958, pruned_loss=0.06854, over 1617678.81 frames. ], batch size: 21, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:27:17,116 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130861.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:27:22,922 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 2.370e+02 2.974e+02 3.638e+02 1.375e+03, threshold=5.949e+02, percent-clipped=1.0 +2023-02-06 19:27:29,124 INFO [train.py:901] (1/4) Epoch 17, batch 1550, loss[loss=0.2188, simple_loss=0.2768, pruned_loss=0.08039, over 7701.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2959, pruned_loss=0.06849, over 1618171.12 frames. ], batch size: 18, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:27:50,132 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130909.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:27:50,701 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130910.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:27:54,296 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130915.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:28:02,600 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:28:03,810 INFO [train.py:901] (1/4) Epoch 17, batch 1600, loss[loss=0.2756, simple_loss=0.3382, pruned_loss=0.1065, over 7056.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.296, pruned_loss=0.06879, over 1611112.33 frames. ], batch size: 71, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:28:34,758 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.345e+02 2.992e+02 3.546e+02 8.486e+02, threshold=5.983e+02, percent-clipped=5.0 +2023-02-06 19:28:40,949 INFO [train.py:901] (1/4) Epoch 17, batch 1650, loss[loss=0.3223, simple_loss=0.3635, pruned_loss=0.1406, over 6982.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2959, pruned_loss=0.06903, over 1609062.51 frames. ], batch size: 73, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:28:57,389 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 19:29:13,495 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131025.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:29:16,139 INFO [train.py:901] (1/4) Epoch 17, batch 1700, loss[loss=0.2264, simple_loss=0.3013, pruned_loss=0.07574, over 8032.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2964, pruned_loss=0.06896, over 1613697.66 frames. ], batch size: 22, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:29:25,394 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:29:46,946 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.451e+02 3.155e+02 3.823e+02 7.811e+02, threshold=6.311e+02, percent-clipped=3.0 +2023-02-06 19:29:53,062 INFO [train.py:901] (1/4) Epoch 17, batch 1750, loss[loss=0.2023, simple_loss=0.2865, pruned_loss=0.05909, over 8472.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2972, pruned_loss=0.06961, over 1610612.48 frames. ], batch size: 27, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:29:58,721 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1028, 1.8126, 2.0625, 1.8840, 1.4336, 1.8395, 2.6766, 2.3860], + device='cuda:1'), covar=tensor([0.0406, 0.1079, 0.1536, 0.1263, 0.0529, 0.1328, 0.0493, 0.0515], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0191, 0.0157, 0.0100, 0.0164, 0.0115, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 19:30:19,608 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131117.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:30:27,910 INFO [train.py:901] (1/4) Epoch 17, batch 1800, loss[loss=0.1883, simple_loss=0.2606, pruned_loss=0.058, over 8235.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2965, pruned_loss=0.0696, over 1610204.25 frames. ], batch size: 22, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:30:37,098 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:30:52,694 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:30:53,338 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2239, 2.1879, 1.6515, 1.9733, 1.7987, 1.4723, 1.6024, 1.6940], + device='cuda:1'), covar=tensor([0.1282, 0.0321, 0.1128, 0.0493, 0.0602, 0.1366, 0.0943, 0.0851], + device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0230, 0.0324, 0.0298, 0.0296, 0.0326, 0.0338, 0.0311], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 19:30:55,957 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.745e+02 3.356e+02 4.683e+02 1.105e+03, threshold=6.712e+02, percent-clipped=11.0 +2023-02-06 19:30:56,918 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131171.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:31:03,697 INFO [train.py:901] (1/4) Epoch 17, batch 1850, loss[loss=0.2361, simple_loss=0.3201, pruned_loss=0.07599, over 8661.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2962, pruned_loss=0.06925, over 1612773.89 frames. ], batch size: 39, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:31:12,470 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131190.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:31:13,824 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131192.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:31:16,714 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131196.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:31:37,448 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0640, 2.7281, 3.6398, 1.9712, 2.0113, 3.5626, 0.8305, 2.3013], + device='cuda:1'), covar=tensor([0.1491, 0.1110, 0.0239, 0.2130, 0.2786, 0.0283, 0.2570, 0.1326], + device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0183, 0.0115, 0.0216, 0.0261, 0.0122, 0.0166, 0.0179], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 19:31:39,956 INFO [train.py:901] (1/4) Epoch 17, batch 1900, loss[loss=0.2047, simple_loss=0.2836, pruned_loss=0.06288, over 8089.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2967, pruned_loss=0.06933, over 1612634.80 frames. ], batch size: 21, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:32:08,105 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.313e+02 2.955e+02 3.582e+02 5.685e+02, threshold=5.910e+02, percent-clipped=0.0 +2023-02-06 19:32:08,141 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 19:32:14,123 INFO [train.py:901] (1/4) Epoch 17, batch 1950, loss[loss=0.2083, simple_loss=0.2987, pruned_loss=0.05897, over 8112.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2964, pruned_loss=0.06896, over 1613190.64 frames. ], batch size: 23, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:32:15,750 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131281.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:32:19,619 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 19:32:28,913 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131298.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:32:35,143 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131306.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:32:39,941 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 19:32:47,535 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131323.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:32:51,200 INFO [train.py:901] (1/4) Epoch 17, batch 2000, loss[loss=0.2541, simple_loss=0.3319, pruned_loss=0.08818, over 8248.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2971, pruned_loss=0.0694, over 1612732.58 frames. ], batch size: 24, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:33:19,856 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.510e+02 3.128e+02 3.622e+02 6.098e+02, threshold=6.257e+02, percent-clipped=1.0 +2023-02-06 19:33:25,355 INFO [train.py:901] (1/4) Epoch 17, batch 2050, loss[loss=0.2047, simple_loss=0.2847, pruned_loss=0.06234, over 8281.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2963, pruned_loss=0.06851, over 1615086.41 frames. ], batch size: 23, lr: 4.51e-03, grad_scale: 8.0 +2023-02-06 19:34:00,659 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131427.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:34:01,963 INFO [train.py:901] (1/4) Epoch 17, batch 2100, loss[loss=0.1678, simple_loss=0.2459, pruned_loss=0.04482, over 7690.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2953, pruned_loss=0.06764, over 1614031.08 frames. ], batch size: 18, lr: 4.51e-03, grad_scale: 8.0 +2023-02-06 19:34:06,118 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:34:31,408 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.457e+02 2.884e+02 3.530e+02 8.686e+02, threshold=5.767e+02, percent-clipped=1.0 +2023-02-06 19:34:36,970 INFO [train.py:901] (1/4) Epoch 17, batch 2150, loss[loss=0.2093, simple_loss=0.271, pruned_loss=0.07378, over 7696.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2956, pruned_loss=0.06792, over 1614480.26 frames. ], batch size: 18, lr: 4.51e-03, grad_scale: 8.0 +2023-02-06 19:34:58,678 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:35:12,357 INFO [train.py:901] (1/4) Epoch 17, batch 2200, loss[loss=0.2243, simple_loss=0.306, pruned_loss=0.0713, over 8337.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2947, pruned_loss=0.06771, over 1612917.97 frames. ], batch size: 26, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:35:17,213 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=131536.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:35:36,120 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3149, 1.9151, 1.4063, 3.0907, 1.4086, 1.2067, 2.0863, 2.1828], + device='cuda:1'), covar=tensor([0.1720, 0.1277, 0.2103, 0.0366, 0.1390, 0.2312, 0.1038, 0.0967], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0198, 0.0247, 0.0211, 0.0208, 0.0246, 0.0253, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 19:35:43,531 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.550e+02 3.248e+02 4.465e+02 1.208e+03, threshold=6.496e+02, percent-clipped=6.0 +2023-02-06 19:35:49,219 INFO [train.py:901] (1/4) Epoch 17, batch 2250, loss[loss=0.2176, simple_loss=0.2811, pruned_loss=0.07704, over 7532.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2961, pruned_loss=0.06854, over 1613427.62 frames. ], batch size: 18, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:36:23,882 INFO [train.py:901] (1/4) Epoch 17, batch 2300, loss[loss=0.2235, simple_loss=0.3003, pruned_loss=0.07337, over 8440.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2965, pruned_loss=0.06879, over 1616756.16 frames. ], batch size: 49, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:36:24,071 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2925, 1.3874, 1.2704, 1.8513, 0.7750, 1.1416, 1.2324, 1.4374], + device='cuda:1'), covar=tensor([0.0894, 0.0917, 0.1119, 0.0508, 0.1172, 0.1491, 0.0851, 0.0822], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0199, 0.0249, 0.0212, 0.0209, 0.0247, 0.0254, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 19:36:40,751 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131651.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:36:55,870 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 2.553e+02 3.001e+02 3.824e+02 6.268e+02, threshold=6.003e+02, percent-clipped=0.0 +2023-02-06 19:37:01,540 INFO [train.py:901] (1/4) Epoch 17, batch 2350, loss[loss=0.2093, simple_loss=0.277, pruned_loss=0.07075, over 7788.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2979, pruned_loss=0.06989, over 1617168.03 frames. ], batch size: 19, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:37:16,046 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8001, 1.9064, 2.4038, 1.6724, 1.3727, 2.3527, 0.4215, 1.4686], + device='cuda:1'), covar=tensor([0.2216, 0.1224, 0.0383, 0.1444, 0.2976, 0.0462, 0.2580, 0.1372], + device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0183, 0.0114, 0.0216, 0.0260, 0.0121, 0.0166, 0.0179], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 19:37:21,843 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-02-06 19:37:35,913 INFO [train.py:901] (1/4) Epoch 17, batch 2400, loss[loss=0.2185, simple_loss=0.2899, pruned_loss=0.07353, over 7923.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2971, pruned_loss=0.06936, over 1613888.64 frames. ], batch size: 20, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:37:40,380 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0907, 3.5153, 2.2091, 2.7125, 2.6804, 2.0601, 2.7926, 2.9555], + device='cuda:1'), covar=tensor([0.1510, 0.0380, 0.1143, 0.0768, 0.0727, 0.1341, 0.0983, 0.1031], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0233, 0.0324, 0.0299, 0.0297, 0.0328, 0.0340, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 19:38:06,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.467e+02 3.155e+02 3.892e+02 8.269e+02, threshold=6.310e+02, percent-clipped=4.0 +2023-02-06 19:38:06,491 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=131771.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:38:12,193 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=131778.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:38:12,830 INFO [train.py:901] (1/4) Epoch 17, batch 2450, loss[loss=0.2299, simple_loss=0.3152, pruned_loss=0.07234, over 8103.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2956, pruned_loss=0.06815, over 1612908.51 frames. ], batch size: 23, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:38:47,881 INFO [train.py:901] (1/4) Epoch 17, batch 2500, loss[loss=0.2086, simple_loss=0.2794, pruned_loss=0.06891, over 7796.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2961, pruned_loss=0.06853, over 1613822.83 frames. ], batch size: 19, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:38:57,326 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 19:39:05,482 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=131854.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:39:17,123 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.481e+02 2.929e+02 3.320e+02 7.417e+02, threshold=5.858e+02, percent-clipped=2.0 +2023-02-06 19:39:20,130 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:39:22,845 INFO [train.py:901] (1/4) Epoch 17, batch 2550, loss[loss=0.1773, simple_loss=0.2538, pruned_loss=0.05038, over 7441.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.295, pruned_loss=0.06824, over 1614618.10 frames. ], batch size: 17, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:39:29,624 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:39:34,619 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131893.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:39:45,579 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131907.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:40:00,843 INFO [train.py:901] (1/4) Epoch 17, batch 2600, loss[loss=0.2076, simple_loss=0.3002, pruned_loss=0.05751, over 8466.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.295, pruned_loss=0.06793, over 1615899.55 frames. ], batch size: 25, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:40:03,027 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131932.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:40:11,295 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7930, 1.9934, 2.2810, 1.2564, 2.3917, 1.6158, 0.7072, 1.9267], + device='cuda:1'), covar=tensor([0.0483, 0.0302, 0.0208, 0.0469, 0.0289, 0.0729, 0.0710, 0.0287], + device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0368, 0.0314, 0.0423, 0.0351, 0.0510, 0.0375, 0.0391], + device='cuda:1'), out_proj_covar=tensor([1.1642e-04, 9.7987e-05, 8.3426e-05, 1.1333e-04, 9.4234e-05, 1.4691e-04, + 1.0232e-04, 1.0501e-04], device='cuda:1') +2023-02-06 19:40:28,773 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:40:29,954 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.407e+02 2.887e+02 3.716e+02 6.826e+02, threshold=5.774e+02, percent-clipped=1.0 +2023-02-06 19:40:35,451 INFO [train.py:901] (1/4) Epoch 17, batch 2650, loss[loss=0.2294, simple_loss=0.3051, pruned_loss=0.07691, over 8465.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2956, pruned_loss=0.0679, over 1618399.02 frames. ], batch size: 27, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:41:09,809 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.50 vs. limit=5.0 +2023-02-06 19:41:13,498 INFO [train.py:901] (1/4) Epoch 17, batch 2700, loss[loss=0.2092, simple_loss=0.2832, pruned_loss=0.06757, over 7817.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2946, pruned_loss=0.06749, over 1611829.04 frames. ], batch size: 20, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:41:27,363 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132049.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:41:42,496 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.450e+02 3.248e+02 4.102e+02 1.137e+03, threshold=6.496e+02, percent-clipped=12.0 +2023-02-06 19:41:43,387 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:41:48,260 INFO [train.py:901] (1/4) Epoch 17, batch 2750, loss[loss=0.2241, simple_loss=0.312, pruned_loss=0.06805, over 8519.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2955, pruned_loss=0.06772, over 1611063.22 frames. ], batch size: 28, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:42:25,035 INFO [train.py:901] (1/4) Epoch 17, batch 2800, loss[loss=0.2181, simple_loss=0.3013, pruned_loss=0.06743, over 8622.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2948, pruned_loss=0.06774, over 1608244.71 frames. ], batch size: 49, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:42:35,384 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132142.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:42:40,216 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132149.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:42:52,683 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132167.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:42:55,272 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.217e+02 2.865e+02 3.623e+02 1.020e+03, threshold=5.730e+02, percent-clipped=3.0 +2023-02-06 19:42:57,349 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132174.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:43:00,623 INFO [train.py:901] (1/4) Epoch 17, batch 2850, loss[loss=0.207, simple_loss=0.2714, pruned_loss=0.07134, over 7197.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2945, pruned_loss=0.06732, over 1611422.40 frames. ], batch size: 16, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:43:11,910 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 19:43:29,198 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=132219.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:43:33,403 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132225.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:43:35,753 INFO [train.py:901] (1/4) Epoch 17, batch 2900, loss[loss=0.2543, simple_loss=0.3299, pruned_loss=0.08939, over 8348.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2951, pruned_loss=0.06846, over 1608762.61 frames. ], batch size: 24, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:43:52,923 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:44:08,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.407e+02 2.887e+02 3.454e+02 7.005e+02, threshold=5.774e+02, percent-clipped=2.0 +2023-02-06 19:44:09,839 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 19:44:13,740 INFO [train.py:901] (1/4) Epoch 17, batch 2950, loss[loss=0.2093, simple_loss=0.2924, pruned_loss=0.06306, over 8361.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2956, pruned_loss=0.06862, over 1608594.68 frames. ], batch size: 24, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:44:36,063 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 19:44:48,313 INFO [train.py:901] (1/4) Epoch 17, batch 3000, loss[loss=0.2068, simple_loss=0.2987, pruned_loss=0.0574, over 8559.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2945, pruned_loss=0.06797, over 1605547.20 frames. ], batch size: 39, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:44:48,313 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 19:44:56,176 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4822, 1.7250, 2.5224, 1.2980, 1.8678, 1.7881, 1.5887, 1.8383], + device='cuda:1'), covar=tensor([0.1787, 0.2571, 0.0953, 0.4475, 0.1848, 0.3214, 0.2115, 0.2293], + device='cuda:1'), in_proj_covar=tensor([0.0509, 0.0566, 0.0546, 0.0619, 0.0634, 0.0574, 0.0510, 0.0623], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:45:00,597 INFO [train.py:935] (1/4) Epoch 17, validation: loss=0.1786, simple_loss=0.2786, pruned_loss=0.03928, over 944034.00 frames. +2023-02-06 19:45:00,598 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 19:45:04,441 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132334.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:45:07,385 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3619, 2.0802, 2.8559, 2.3292, 2.7271, 2.3587, 2.0359, 1.4779], + device='cuda:1'), covar=tensor([0.4640, 0.4663, 0.1590, 0.3297, 0.2295, 0.2717, 0.1747, 0.5013], + device='cuda:1'), in_proj_covar=tensor([0.0931, 0.0938, 0.0776, 0.0908, 0.0973, 0.0858, 0.0727, 0.0809], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 19:45:12,171 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3994, 1.6110, 1.7984, 1.4987, 1.0645, 1.4834, 1.9701, 1.7962], + device='cuda:1'), covar=tensor([0.0456, 0.1224, 0.1604, 0.1375, 0.0564, 0.1505, 0.0621, 0.0612], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0191, 0.0157, 0.0100, 0.0163, 0.0115, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 19:45:31,448 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.547e+02 2.492e+02 3.005e+02 3.786e+02 8.313e+02, threshold=6.010e+02, percent-clipped=11.0 +2023-02-06 19:45:37,100 INFO [train.py:901] (1/4) Epoch 17, batch 3050, loss[loss=0.2174, simple_loss=0.304, pruned_loss=0.06542, over 8459.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2957, pruned_loss=0.06819, over 1610812.38 frames. ], batch size: 27, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:45:48,262 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=132393.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:46:04,203 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=132416.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:46:12,931 INFO [train.py:901] (1/4) Epoch 17, batch 3100, loss[loss=0.1798, simple_loss=0.2586, pruned_loss=0.05056, over 7795.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2958, pruned_loss=0.0679, over 1614023.36 frames. ], batch size: 19, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:46:41,880 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.340e+02 2.843e+02 3.195e+02 7.960e+02, threshold=5.685e+02, percent-clipped=6.0 +2023-02-06 19:46:47,319 INFO [train.py:901] (1/4) Epoch 17, batch 3150, loss[loss=0.2682, simple_loss=0.3372, pruned_loss=0.09954, over 8335.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2955, pruned_loss=0.06796, over 1612264.21 frames. ], batch size: 26, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:46:48,918 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5021, 2.6745, 2.0309, 2.3952, 2.4577, 1.7388, 2.3102, 2.3163], + device='cuda:1'), covar=tensor([0.1458, 0.0383, 0.1040, 0.0563, 0.0680, 0.1369, 0.0828, 0.0910], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0233, 0.0324, 0.0300, 0.0297, 0.0328, 0.0339, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 19:47:09,719 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132508.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:47:23,941 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.57 vs. limit=5.0 +2023-02-06 19:47:24,992 INFO [train.py:901] (1/4) Epoch 17, batch 3200, loss[loss=0.2058, simple_loss=0.297, pruned_loss=0.05729, over 8513.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2944, pruned_loss=0.06748, over 1613597.19 frames. ], batch size: 28, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:47:26,580 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:47:54,169 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.495e+02 3.112e+02 3.824e+02 1.248e+03, threshold=6.223e+02, percent-clipped=6.0 +2023-02-06 19:47:59,503 INFO [train.py:901] (1/4) Epoch 17, batch 3250, loss[loss=0.2327, simple_loss=0.31, pruned_loss=0.07775, over 8419.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2952, pruned_loss=0.06778, over 1611667.37 frames. ], batch size: 49, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:48:07,396 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132590.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:48:11,622 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4679, 2.5738, 1.8606, 2.3269, 2.3317, 1.5909, 2.1260, 2.2284], + device='cuda:1'), covar=tensor([0.1694, 0.0380, 0.1170, 0.0627, 0.0673, 0.1583, 0.0938, 0.0965], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0233, 0.0324, 0.0301, 0.0298, 0.0330, 0.0341, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 19:48:26,285 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132615.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:48:37,513 INFO [train.py:901] (1/4) Epoch 17, batch 3300, loss[loss=0.2445, simple_loss=0.3133, pruned_loss=0.08782, over 6741.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.296, pruned_loss=0.06789, over 1611193.04 frames. ], batch size: 71, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:49:06,785 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.345e+02 2.532e+02 2.971e+02 3.744e+02 7.972e+02, threshold=5.942e+02, percent-clipped=3.0 +2023-02-06 19:49:11,875 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132678.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:49:12,450 INFO [train.py:901] (1/4) Epoch 17, batch 3350, loss[loss=0.1703, simple_loss=0.2595, pruned_loss=0.04058, over 8292.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2957, pruned_loss=0.06725, over 1614501.74 frames. ], batch size: 23, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:49:49,256 INFO [train.py:901] (1/4) Epoch 17, batch 3400, loss[loss=0.2393, simple_loss=0.3138, pruned_loss=0.08235, over 8498.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.296, pruned_loss=0.0675, over 1612661.66 frames. ], batch size: 28, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:49:55,919 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6703, 5.7494, 5.0868, 2.4586, 5.1339, 5.5306, 5.2900, 5.1772], + device='cuda:1'), covar=tensor([0.0535, 0.0352, 0.0901, 0.4551, 0.0699, 0.0771, 0.0998, 0.0596], + device='cuda:1'), in_proj_covar=tensor([0.0501, 0.0414, 0.0421, 0.0513, 0.0406, 0.0414, 0.0400, 0.0358], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:50:01,001 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 19:50:04,442 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.21 vs. limit=5.0 +2023-02-06 19:50:14,686 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132764.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:14,765 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132764.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:19,431 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.442e+02 2.969e+02 4.012e+02 9.663e+02, threshold=5.937e+02, percent-clipped=5.0 +2023-02-06 19:50:24,937 INFO [train.py:901] (1/4) Epoch 17, batch 3450, loss[loss=0.2242, simple_loss=0.3089, pruned_loss=0.0698, over 7636.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2962, pruned_loss=0.06763, over 1616361.31 frames. ], batch size: 19, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:50:30,900 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132787.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:32,218 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:33,515 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5907, 1.9341, 1.9350, 1.2280, 1.9971, 1.5072, 0.4686, 1.7573], + device='cuda:1'), covar=tensor([0.0412, 0.0296, 0.0195, 0.0429, 0.0351, 0.0727, 0.0757, 0.0236], + device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0370, 0.0315, 0.0427, 0.0353, 0.0510, 0.0378, 0.0392], + device='cuda:1'), out_proj_covar=tensor([1.1769e-04, 9.8406e-05, 8.3563e-05, 1.1444e-04, 9.4984e-05, 1.4697e-04, + 1.0310e-04, 1.0527e-04], device='cuda:1') +2023-02-06 19:50:47,676 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132812.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:59,803 INFO [train.py:901] (1/4) Epoch 17, batch 3500, loss[loss=0.2083, simple_loss=0.2889, pruned_loss=0.06385, over 8089.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2968, pruned_loss=0.06834, over 1616711.43 frames. ], batch size: 21, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:51:13,854 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 19:51:31,535 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.520e+02 2.401e+02 3.009e+02 3.970e+02 8.620e+02, threshold=6.019e+02, percent-clipped=6.0 +2023-02-06 19:51:37,017 INFO [train.py:901] (1/4) Epoch 17, batch 3550, loss[loss=0.2204, simple_loss=0.3042, pruned_loss=0.06827, over 8443.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2968, pruned_loss=0.06825, over 1617122.60 frames. ], batch size: 29, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:52:11,133 INFO [train.py:901] (1/4) Epoch 17, batch 3600, loss[loss=0.1625, simple_loss=0.2483, pruned_loss=0.03838, over 7539.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2956, pruned_loss=0.06765, over 1613899.32 frames. ], batch size: 18, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:52:11,368 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9482, 1.1439, 1.0851, 0.5664, 1.1304, 0.9269, 0.0580, 1.1089], + device='cuda:1'), covar=tensor([0.0333, 0.0314, 0.0268, 0.0450, 0.0322, 0.0784, 0.0652, 0.0256], + device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0375, 0.0319, 0.0431, 0.0358, 0.0515, 0.0380, 0.0396], + device='cuda:1'), out_proj_covar=tensor([1.1865e-04, 9.9762e-05, 8.4682e-05, 1.1547e-04, 9.6214e-05, 1.4857e-04, + 1.0381e-04, 1.0618e-04], device='cuda:1') +2023-02-06 19:52:40,001 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8518, 1.9765, 2.0817, 1.5938, 2.1656, 1.5787, 0.9828, 1.9127], + device='cuda:1'), covar=tensor([0.0478, 0.0291, 0.0245, 0.0422, 0.0325, 0.0672, 0.0704, 0.0269], + device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0373, 0.0318, 0.0429, 0.0356, 0.0512, 0.0378, 0.0394], + device='cuda:1'), out_proj_covar=tensor([1.1802e-04, 9.9351e-05, 8.4339e-05, 1.1490e-04, 9.5670e-05, 1.4765e-04, + 1.0323e-04, 1.0562e-04], device='cuda:1') +2023-02-06 19:52:41,874 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.442e+02 2.775e+02 3.418e+02 6.006e+02, threshold=5.549e+02, percent-clipped=0.0 +2023-02-06 19:52:48,328 INFO [train.py:901] (1/4) Epoch 17, batch 3650, loss[loss=0.2144, simple_loss=0.2769, pruned_loss=0.07599, over 7418.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2946, pruned_loss=0.06712, over 1614999.53 frames. ], batch size: 17, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:53:18,534 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:53:21,775 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 19:53:23,086 INFO [train.py:901] (1/4) Epoch 17, batch 3700, loss[loss=0.2306, simple_loss=0.2999, pruned_loss=0.08063, over 8075.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2934, pruned_loss=0.06655, over 1614310.80 frames. ], batch size: 21, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:53:37,039 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 19:53:53,567 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.419e+02 3.081e+02 4.194e+02 7.364e+02, threshold=6.162e+02, percent-clipped=6.0 +2023-02-06 19:53:59,117 INFO [train.py:901] (1/4) Epoch 17, batch 3750, loss[loss=0.2069, simple_loss=0.2858, pruned_loss=0.064, over 7665.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.294, pruned_loss=0.06671, over 1612834.27 frames. ], batch size: 19, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:54:10,363 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8229, 5.9704, 5.1321, 2.5538, 5.2198, 5.6338, 5.3933, 5.3950], + device='cuda:1'), covar=tensor([0.0555, 0.0371, 0.0915, 0.4484, 0.0758, 0.0664, 0.1104, 0.0581], + device='cuda:1'), in_proj_covar=tensor([0.0504, 0.0412, 0.0419, 0.0514, 0.0408, 0.0416, 0.0400, 0.0360], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 19:54:21,510 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133108.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:54:35,500 INFO [train.py:901] (1/4) Epoch 17, batch 3800, loss[loss=0.2236, simple_loss=0.3047, pruned_loss=0.07128, over 8664.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2947, pruned_loss=0.06689, over 1615313.00 frames. ], batch size: 34, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:54:41,265 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133137.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:55:04,561 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.594e+02 3.054e+02 3.718e+02 6.772e+02, threshold=6.108e+02, percent-clipped=5.0 +2023-02-06 19:55:09,936 INFO [train.py:901] (1/4) Epoch 17, batch 3850, loss[loss=0.1968, simple_loss=0.2783, pruned_loss=0.05765, over 7931.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2955, pruned_loss=0.06704, over 1619083.14 frames. ], batch size: 20, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:55:31,162 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 19:55:39,010 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4357, 2.3424, 3.1745, 2.5177, 2.8576, 2.4225, 2.1848, 1.7699], + device='cuda:1'), covar=tensor([0.4650, 0.4539, 0.1676, 0.3397, 0.2635, 0.2695, 0.1739, 0.5108], + device='cuda:1'), in_proj_covar=tensor([0.0918, 0.0931, 0.0770, 0.0902, 0.0968, 0.0850, 0.0721, 0.0800], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 19:55:43,012 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133223.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:55:46,904 INFO [train.py:901] (1/4) Epoch 17, batch 3900, loss[loss=0.2027, simple_loss=0.2848, pruned_loss=0.0603, over 7811.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2946, pruned_loss=0.06687, over 1619853.36 frames. ], batch size: 20, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:56:15,757 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.486e+02 2.968e+02 4.028e+02 1.073e+03, threshold=5.936e+02, percent-clipped=5.0 +2023-02-06 19:56:21,114 INFO [train.py:901] (1/4) Epoch 17, batch 3950, loss[loss=0.1768, simple_loss=0.2616, pruned_loss=0.04605, over 7436.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2942, pruned_loss=0.06658, over 1616570.37 frames. ], batch size: 17, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:56:53,890 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.14 vs. limit=5.0 +2023-02-06 19:56:56,946 INFO [train.py:901] (1/4) Epoch 17, batch 4000, loss[loss=0.2016, simple_loss=0.2829, pruned_loss=0.06013, over 8191.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2939, pruned_loss=0.06662, over 1611492.41 frames. ], batch size: 23, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:57:27,413 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.526e+02 3.333e+02 3.995e+02 7.649e+02, threshold=6.666e+02, percent-clipped=5.0 +2023-02-06 19:57:32,336 INFO [train.py:901] (1/4) Epoch 17, batch 4050, loss[loss=0.1994, simple_loss=0.2786, pruned_loss=0.06007, over 8562.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2942, pruned_loss=0.06711, over 1606000.18 frames. ], batch size: 31, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:57:38,097 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 19:57:41,370 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133392.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:57:42,105 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133393.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:57:42,763 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133394.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:57:59,642 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133418.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:58:07,778 INFO [train.py:901] (1/4) Epoch 17, batch 4100, loss[loss=0.2221, simple_loss=0.311, pruned_loss=0.06658, over 8535.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2936, pruned_loss=0.0667, over 1609509.17 frames. ], batch size: 31, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:58:40,076 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 2.513e+02 2.919e+02 3.658e+02 1.440e+03, threshold=5.839e+02, percent-clipped=2.0 +2023-02-06 19:58:45,042 INFO [train.py:901] (1/4) Epoch 17, batch 4150, loss[loss=0.1694, simple_loss=0.2443, pruned_loss=0.04725, over 7549.00 frames. ], tot_loss[loss=0.214, simple_loss=0.294, pruned_loss=0.06705, over 1604781.89 frames. ], batch size: 18, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:58:45,257 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:58:54,024 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133492.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:59:00,147 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4794, 2.4096, 1.5668, 2.2612, 2.1268, 1.3679, 2.0528, 2.1384], + device='cuda:1'), covar=tensor([0.1389, 0.0458, 0.1387, 0.0604, 0.0736, 0.1716, 0.0918, 0.0907], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0231, 0.0321, 0.0298, 0.0296, 0.0327, 0.0338, 0.0312], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 19:59:02,283 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:59:19,506 INFO [train.py:901] (1/4) Epoch 17, batch 4200, loss[loss=0.2079, simple_loss=0.2928, pruned_loss=0.06151, over 8509.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.295, pruned_loss=0.06718, over 1605458.02 frames. ], batch size: 49, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:59:32,483 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 19:59:51,070 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.728e+02 2.565e+02 3.135e+02 3.827e+02 1.180e+03, threshold=6.269e+02, percent-clipped=6.0 +2023-02-06 19:59:56,756 INFO [train.py:901] (1/4) Epoch 17, batch 4250, loss[loss=0.198, simple_loss=0.2917, pruned_loss=0.05215, over 8321.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2954, pruned_loss=0.06747, over 1603915.70 frames. ], batch size: 26, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:59:57,436 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 20:00:22,922 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2940, 4.1936, 3.8933, 1.9848, 3.9142, 3.8380, 3.7954, 3.6113], + device='cuda:1'), covar=tensor([0.0687, 0.0551, 0.1035, 0.4200, 0.0818, 0.0855, 0.1395, 0.0714], + device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0408, 0.0417, 0.0509, 0.0404, 0.0408, 0.0397, 0.0356], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:00:30,986 INFO [train.py:901] (1/4) Epoch 17, batch 4300, loss[loss=0.1918, simple_loss=0.278, pruned_loss=0.0528, over 8139.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2939, pruned_loss=0.0669, over 1607074.81 frames. ], batch size: 22, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:01:00,707 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133670.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:01:01,926 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.551e+02 3.118e+02 3.976e+02 6.360e+02, threshold=6.236e+02, percent-clipped=1.0 +2023-02-06 20:01:06,893 INFO [train.py:901] (1/4) Epoch 17, batch 4350, loss[loss=0.228, simple_loss=0.3121, pruned_loss=0.07198, over 8660.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2936, pruned_loss=0.06687, over 1605576.77 frames. ], batch size: 34, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:01:31,278 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 20:01:43,144 INFO [train.py:901] (1/4) Epoch 17, batch 4400, loss[loss=0.1843, simple_loss=0.2687, pruned_loss=0.04997, over 7798.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2919, pruned_loss=0.06604, over 1603436.51 frames. ], batch size: 19, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:01:48,111 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133736.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 20:01:49,429 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133738.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:02:12,877 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.579e+02 3.148e+02 3.884e+02 8.584e+02, threshold=6.297e+02, percent-clipped=6.0 +2023-02-06 20:02:12,923 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 20:02:18,530 INFO [train.py:901] (1/4) Epoch 17, batch 4450, loss[loss=0.2056, simple_loss=0.279, pruned_loss=0.06614, over 7798.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2911, pruned_loss=0.06533, over 1605232.79 frames. ], batch size: 20, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:02:25,362 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8938, 5.9240, 5.2155, 2.4641, 5.3792, 5.6782, 5.4259, 5.3754], + device='cuda:1'), covar=tensor([0.0540, 0.0396, 0.0921, 0.4583, 0.0650, 0.0680, 0.1008, 0.0474], + device='cuda:1'), in_proj_covar=tensor([0.0503, 0.0412, 0.0421, 0.0515, 0.0406, 0.0413, 0.0400, 0.0359], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:02:55,023 INFO [train.py:901] (1/4) Epoch 17, batch 4500, loss[loss=0.2083, simple_loss=0.283, pruned_loss=0.06685, over 8143.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2918, pruned_loss=0.06596, over 1606647.08 frames. ], batch size: 22, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:03:00,076 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133836.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:03:10,420 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133851.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 20:03:10,902 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 20:03:11,728 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133853.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:03:24,314 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.330e+02 2.856e+02 3.592e+02 8.327e+02, threshold=5.711e+02, percent-clipped=1.0 +2023-02-06 20:03:29,184 INFO [train.py:901] (1/4) Epoch 17, batch 4550, loss[loss=0.2601, simple_loss=0.3175, pruned_loss=0.1013, over 6863.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2938, pruned_loss=0.06719, over 1606936.30 frames. ], batch size: 71, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:04:04,548 INFO [train.py:901] (1/4) Epoch 17, batch 4600, loss[loss=0.2623, simple_loss=0.3126, pruned_loss=0.106, over 7539.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2932, pruned_loss=0.06691, over 1608163.88 frames. ], batch size: 18, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:04:21,357 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:04:23,721 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 20:04:35,425 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.386e+02 2.834e+02 3.772e+02 7.696e+02, threshold=5.668e+02, percent-clipped=3.0 +2023-02-06 20:04:40,245 INFO [train.py:901] (1/4) Epoch 17, batch 4650, loss[loss=0.2585, simple_loss=0.3352, pruned_loss=0.09086, over 8360.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2941, pruned_loss=0.06737, over 1613246.31 frames. ], batch size: 26, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:04:50,869 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1481, 1.4046, 4.3301, 1.5705, 3.8249, 3.5559, 3.8788, 3.7432], + device='cuda:1'), covar=tensor([0.0565, 0.4725, 0.0579, 0.4176, 0.1139, 0.1020, 0.0639, 0.0745], + device='cuda:1'), in_proj_covar=tensor([0.0575, 0.0618, 0.0661, 0.0591, 0.0672, 0.0580, 0.0572, 0.0636], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 20:04:50,938 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3037, 2.4180, 1.7274, 2.0848, 2.0128, 1.4502, 1.8823, 1.9269], + device='cuda:1'), covar=tensor([0.1641, 0.0411, 0.1160, 0.0622, 0.0698, 0.1564, 0.1025, 0.0967], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0233, 0.0325, 0.0301, 0.0297, 0.0331, 0.0341, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 20:05:00,934 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1888, 1.6778, 4.4126, 1.7874, 2.3133, 5.0437, 5.0461, 4.3208], + device='cuda:1'), covar=tensor([0.1235, 0.1858, 0.0289, 0.2150, 0.1308, 0.0166, 0.0363, 0.0543], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0311, 0.0275, 0.0304, 0.0295, 0.0253, 0.0392, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 20:05:06,470 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=134014.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:05:07,217 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134015.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:05:13,074 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.02 vs. limit=5.0 +2023-02-06 20:05:16,660 INFO [train.py:901] (1/4) Epoch 17, batch 4700, loss[loss=0.1664, simple_loss=0.2504, pruned_loss=0.04117, over 7565.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2942, pruned_loss=0.0675, over 1612157.01 frames. ], batch size: 18, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:05:48,978 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.331e+02 2.674e+02 3.349e+02 6.559e+02, threshold=5.348e+02, percent-clipped=3.0 +2023-02-06 20:05:53,953 INFO [train.py:901] (1/4) Epoch 17, batch 4750, loss[loss=0.2243, simple_loss=0.3009, pruned_loss=0.07389, over 8511.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2932, pruned_loss=0.06683, over 1609739.85 frames. ], batch size: 26, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:06:13,290 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134107.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 20:06:13,882 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4872, 1.1683, 4.6890, 1.8605, 4.1273, 3.8702, 4.1790, 4.0778], + device='cuda:1'), covar=tensor([0.0663, 0.5206, 0.0537, 0.3796, 0.1158, 0.1000, 0.0655, 0.0736], + device='cuda:1'), in_proj_covar=tensor([0.0576, 0.0620, 0.0662, 0.0591, 0.0675, 0.0580, 0.0575, 0.0640], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 20:06:14,652 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134109.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:06:17,884 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 20:06:20,576 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 20:06:28,172 INFO [train.py:901] (1/4) Epoch 17, batch 4800, loss[loss=0.2228, simple_loss=0.3087, pruned_loss=0.06843, over 8190.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2943, pruned_loss=0.06695, over 1615446.88 frames. ], batch size: 23, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:06:28,371 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134129.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:06:31,287 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134132.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 20:06:32,686 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:06:38,305 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3076, 1.7400, 1.8659, 1.1220, 1.8831, 1.2499, 0.3453, 1.5303], + device='cuda:1'), covar=tensor([0.0576, 0.0357, 0.0256, 0.0527, 0.0398, 0.0949, 0.0832, 0.0297], + device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0372, 0.0316, 0.0427, 0.0354, 0.0512, 0.0377, 0.0395], + device='cuda:1'), out_proj_covar=tensor([1.1799e-04, 9.9152e-05, 8.3567e-05, 1.1411e-04, 9.4922e-05, 1.4763e-04, + 1.0279e-04, 1.0596e-04], device='cuda:1') +2023-02-06 20:06:46,856 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3529, 1.7585, 1.8724, 1.1307, 1.8524, 1.3307, 0.3736, 1.5542], + device='cuda:1'), covar=tensor([0.0607, 0.0398, 0.0320, 0.0562, 0.0476, 0.0878, 0.0854, 0.0325], + device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0373, 0.0317, 0.0429, 0.0355, 0.0514, 0.0379, 0.0397], + device='cuda:1'), out_proj_covar=tensor([1.1836e-04, 9.9603e-05, 8.3850e-05, 1.1452e-04, 9.5190e-05, 1.4809e-04, + 1.0316e-04, 1.0638e-04], device='cuda:1') +2023-02-06 20:07:00,736 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.352e+02 2.869e+02 3.488e+02 8.440e+02, threshold=5.739e+02, percent-clipped=9.0 +2023-02-06 20:07:06,353 INFO [train.py:901] (1/4) Epoch 17, batch 4850, loss[loss=0.2584, simple_loss=0.3343, pruned_loss=0.09126, over 8464.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.293, pruned_loss=0.06588, over 1616638.54 frames. ], batch size: 29, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:07:14,643 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 20:07:26,239 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134207.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:07:41,130 INFO [train.py:901] (1/4) Epoch 17, batch 4900, loss[loss=0.2173, simple_loss=0.3093, pruned_loss=0.06262, over 8246.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2948, pruned_loss=0.0668, over 1617090.19 frames. ], batch size: 24, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:07:43,518 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134232.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:08:10,409 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3392, 1.7366, 4.0591, 1.9587, 2.6192, 4.5696, 4.6424, 3.9444], + device='cuda:1'), covar=tensor([0.1056, 0.1773, 0.0393, 0.1927, 0.1250, 0.0195, 0.0391, 0.0570], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0310, 0.0275, 0.0302, 0.0294, 0.0254, 0.0390, 0.0298], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 20:08:11,581 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-06 20:08:13,115 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.657e+02 3.351e+02 4.707e+02 1.168e+03, threshold=6.701e+02, percent-clipped=12.0 +2023-02-06 20:08:17,821 INFO [train.py:901] (1/4) Epoch 17, batch 4950, loss[loss=0.2113, simple_loss=0.2978, pruned_loss=0.06234, over 8717.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2942, pruned_loss=0.0665, over 1616881.87 frames. ], batch size: 39, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:08:18,374 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.98 vs. limit=5.0 +2023-02-06 20:08:49,557 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:08:52,968 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:08:54,220 INFO [train.py:901] (1/4) Epoch 17, batch 5000, loss[loss=0.2158, simple_loss=0.2976, pruned_loss=0.06701, over 8239.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2945, pruned_loss=0.0666, over 1617879.12 frames. ], batch size: 22, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:09:15,211 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=134359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:09:24,842 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.361e+02 2.656e+02 3.405e+02 6.362e+02, threshold=5.311e+02, percent-clipped=0.0 +2023-02-06 20:09:30,487 INFO [train.py:901] (1/4) Epoch 17, batch 5050, loss[loss=0.2054, simple_loss=0.2966, pruned_loss=0.05708, over 8136.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2962, pruned_loss=0.06824, over 1619927.20 frames. ], batch size: 22, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:09:32,845 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3389, 1.3711, 4.5641, 1.6624, 4.0036, 3.8318, 4.1389, 3.9733], + device='cuda:1'), covar=tensor([0.0598, 0.4484, 0.0487, 0.3729, 0.1103, 0.0910, 0.0559, 0.0699], + device='cuda:1'), in_proj_covar=tensor([0.0578, 0.0620, 0.0664, 0.0591, 0.0676, 0.0580, 0.0575, 0.0640], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 20:09:35,086 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134385.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:09:54,075 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134410.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:09:58,691 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 20:10:07,172 INFO [train.py:901] (1/4) Epoch 17, batch 5100, loss[loss=0.2338, simple_loss=0.3052, pruned_loss=0.08121, over 8651.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2948, pruned_loss=0.06739, over 1619307.90 frames. ], batch size: 34, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:10:36,990 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.440e+02 2.384e+02 2.769e+02 3.675e+02 1.185e+03, threshold=5.538e+02, percent-clipped=9.0 +2023-02-06 20:10:38,540 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134474.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:10:42,687 INFO [train.py:901] (1/4) Epoch 17, batch 5150, loss[loss=0.2158, simple_loss=0.2932, pruned_loss=0.06915, over 7925.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2943, pruned_loss=0.06742, over 1614024.93 frames. ], batch size: 20, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:10:45,525 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8010, 5.9311, 4.9876, 2.4901, 5.1606, 5.5677, 5.4152, 5.3318], + device='cuda:1'), covar=tensor([0.0473, 0.0367, 0.0948, 0.4508, 0.0622, 0.0585, 0.0975, 0.0458], + device='cuda:1'), in_proj_covar=tensor([0.0500, 0.0412, 0.0418, 0.0514, 0.0404, 0.0412, 0.0399, 0.0359], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:11:20,277 INFO [train.py:901] (1/4) Epoch 17, batch 5200, loss[loss=0.2078, simple_loss=0.2857, pruned_loss=0.06494, over 7694.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2927, pruned_loss=0.06654, over 1613426.16 frames. ], batch size: 18, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:11:49,976 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.225e+02 2.783e+02 3.706e+02 1.482e+03, threshold=5.567e+02, percent-clipped=8.0 +2023-02-06 20:11:54,882 INFO [train.py:901] (1/4) Epoch 17, batch 5250, loss[loss=0.198, simple_loss=0.2783, pruned_loss=0.05891, over 8229.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2937, pruned_loss=0.06736, over 1614528.65 frames. ], batch size: 22, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:11:57,586 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 20:12:31,033 INFO [train.py:901] (1/4) Epoch 17, batch 5300, loss[loss=0.2284, simple_loss=0.3058, pruned_loss=0.07548, over 8134.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2943, pruned_loss=0.06781, over 1615533.47 frames. ], batch size: 22, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:12:48,113 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5583, 1.9730, 2.0479, 1.1004, 2.1486, 1.4640, 0.4797, 1.8364], + device='cuda:1'), covar=tensor([0.0570, 0.0279, 0.0231, 0.0573, 0.0345, 0.0851, 0.0742, 0.0246], + device='cuda:1'), in_proj_covar=tensor([0.0438, 0.0377, 0.0320, 0.0433, 0.0361, 0.0522, 0.0381, 0.0398], + device='cuda:1'), out_proj_covar=tensor([1.1968e-04, 1.0055e-04, 8.4848e-05, 1.1596e-04, 9.6692e-05, 1.5060e-04, + 1.0387e-04, 1.0658e-04], device='cuda:1') +2023-02-06 20:12:58,383 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=134666.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:13:01,795 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=134671.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:13:02,357 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 2.333e+02 2.884e+02 3.429e+02 1.143e+03, threshold=5.769e+02, percent-clipped=6.0 +2023-02-06 20:13:07,135 INFO [train.py:901] (1/4) Epoch 17, batch 5350, loss[loss=0.2161, simple_loss=0.2973, pruned_loss=0.06742, over 8332.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2948, pruned_loss=0.06764, over 1614059.31 frames. ], batch size: 26, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:13:43,348 INFO [train.py:901] (1/4) Epoch 17, batch 5400, loss[loss=0.2275, simple_loss=0.3007, pruned_loss=0.07712, over 7648.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2961, pruned_loss=0.06839, over 1618606.79 frames. ], batch size: 19, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:13:44,283 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134730.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:14:01,965 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:14:14,300 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.337e+02 2.988e+02 3.635e+02 1.067e+03, threshold=5.976e+02, percent-clipped=7.0 +2023-02-06 20:14:18,980 INFO [train.py:901] (1/4) Epoch 17, batch 5450, loss[loss=0.2524, simple_loss=0.3328, pruned_loss=0.08598, over 8292.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2956, pruned_loss=0.06796, over 1615331.75 frames. ], batch size: 23, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:14:20,483 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134781.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:14:23,983 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134786.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:14:25,739 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-02-06 20:14:47,256 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-02-06 20:14:52,690 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5791, 1.3445, 1.6037, 1.2586, 0.8988, 1.3641, 1.5878, 1.3394], + device='cuda:1'), covar=tensor([0.0524, 0.1304, 0.1697, 0.1477, 0.0606, 0.1614, 0.0691, 0.0658], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0189, 0.0156, 0.0100, 0.0162, 0.0114, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 20:14:54,786 INFO [train.py:901] (1/4) Epoch 17, batch 5500, loss[loss=0.1901, simple_loss=0.2679, pruned_loss=0.05616, over 7653.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2953, pruned_loss=0.06756, over 1618128.04 frames. ], batch size: 19, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:14:55,412 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 20:15:25,527 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.381e+02 2.895e+02 3.783e+02 8.489e+02, threshold=5.790e+02, percent-clipped=3.0 +2023-02-06 20:15:31,374 INFO [train.py:901] (1/4) Epoch 17, batch 5550, loss[loss=0.2286, simple_loss=0.3145, pruned_loss=0.07137, over 8392.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2952, pruned_loss=0.0672, over 1620092.42 frames. ], batch size: 49, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:15:32,210 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134880.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:15:42,718 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3545, 1.4886, 2.1747, 1.2226, 1.4211, 1.6290, 1.3725, 1.4900], + device='cuda:1'), covar=tensor([0.1857, 0.2557, 0.0902, 0.4440, 0.1915, 0.3172, 0.2287, 0.2133], + device='cuda:1'), in_proj_covar=tensor([0.0509, 0.0567, 0.0544, 0.0615, 0.0633, 0.0572, 0.0508, 0.0619], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:16:06,822 INFO [train.py:901] (1/4) Epoch 17, batch 5600, loss[loss=0.2736, simple_loss=0.3541, pruned_loss=0.09655, over 8587.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2945, pruned_loss=0.0668, over 1617840.23 frames. ], batch size: 34, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:16:38,754 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.724e+02 3.292e+02 4.135e+02 9.276e+02, threshold=6.584e+02, percent-clipped=7.0 +2023-02-06 20:16:42,892 INFO [train.py:901] (1/4) Epoch 17, batch 5650, loss[loss=0.2075, simple_loss=0.2992, pruned_loss=0.05783, over 7819.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2954, pruned_loss=0.06764, over 1616690.86 frames. ], batch size: 20, lr: 4.45e-03, grad_scale: 4.0 +2023-02-06 20:16:47,882 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7601, 1.4807, 5.8831, 2.2159, 5.2478, 4.8570, 5.4407, 5.2896], + device='cuda:1'), covar=tensor([0.0488, 0.5006, 0.0340, 0.3598, 0.0988, 0.0815, 0.0512, 0.0523], + device='cuda:1'), in_proj_covar=tensor([0.0587, 0.0628, 0.0670, 0.0601, 0.0680, 0.0584, 0.0579, 0.0647], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:17:04,362 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 20:17:18,735 INFO [train.py:901] (1/4) Epoch 17, batch 5700, loss[loss=0.2067, simple_loss=0.2796, pruned_loss=0.06687, over 7965.00 frames. ], tot_loss[loss=0.215, simple_loss=0.295, pruned_loss=0.06749, over 1615738.51 frames. ], batch size: 21, lr: 4.45e-03, grad_scale: 4.0 +2023-02-06 20:17:24,526 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135037.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:17:27,998 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:17:42,576 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:17:45,934 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135067.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:17:49,723 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.516e+02 3.214e+02 3.973e+02 1.283e+03, threshold=6.427e+02, percent-clipped=6.0 +2023-02-06 20:17:53,703 INFO [train.py:901] (1/4) Epoch 17, batch 5750, loss[loss=0.2403, simple_loss=0.3118, pruned_loss=0.08441, over 8292.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2948, pruned_loss=0.06745, over 1620054.42 frames. ], batch size: 23, lr: 4.45e-03, grad_scale: 4.0 +2023-02-06 20:18:11,549 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 20:18:14,104 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-02-06 20:18:30,180 INFO [train.py:901] (1/4) Epoch 17, batch 5800, loss[loss=0.2647, simple_loss=0.3488, pruned_loss=0.09029, over 8654.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2947, pruned_loss=0.06702, over 1619714.49 frames. ], batch size: 39, lr: 4.44e-03, grad_scale: 4.0 +2023-02-06 20:19:00,425 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.84 vs. limit=5.0 +2023-02-06 20:19:00,545 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.313e+02 2.882e+02 3.681e+02 6.576e+02, threshold=5.764e+02, percent-clipped=1.0 +2023-02-06 20:19:04,575 INFO [train.py:901] (1/4) Epoch 17, batch 5850, loss[loss=0.2638, simple_loss=0.3123, pruned_loss=0.1077, over 7703.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2946, pruned_loss=0.06701, over 1619353.12 frames. ], batch size: 18, lr: 4.44e-03, grad_scale: 4.0 +2023-02-06 20:19:12,155 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135189.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:19:37,622 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=135224.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:19:41,091 INFO [train.py:901] (1/4) Epoch 17, batch 5900, loss[loss=0.1754, simple_loss=0.2596, pruned_loss=0.04562, over 7810.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2949, pruned_loss=0.06731, over 1619111.43 frames. ], batch size: 19, lr: 4.44e-03, grad_scale: 4.0 +2023-02-06 20:20:12,437 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.357e+02 3.084e+02 3.660e+02 6.807e+02, threshold=6.167e+02, percent-clipped=2.0 +2023-02-06 20:20:16,629 INFO [train.py:901] (1/4) Epoch 17, batch 5950, loss[loss=0.1777, simple_loss=0.2453, pruned_loss=0.05504, over 7683.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2947, pruned_loss=0.06726, over 1620845.37 frames. ], batch size: 18, lr: 4.44e-03, grad_scale: 4.0 +2023-02-06 20:20:52,280 INFO [train.py:901] (1/4) Epoch 17, batch 6000, loss[loss=0.2237, simple_loss=0.2998, pruned_loss=0.07385, over 7814.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2952, pruned_loss=0.06833, over 1615581.63 frames. ], batch size: 20, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:20:52,280 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 20:21:02,381 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7755, 3.7388, 3.4645, 2.1090, 3.3343, 3.4206, 3.4460, 3.1698], + device='cuda:1'), covar=tensor([0.0981, 0.0552, 0.0920, 0.4684, 0.1013, 0.0920, 0.1166, 0.0918], + device='cuda:1'), in_proj_covar=tensor([0.0501, 0.0412, 0.0417, 0.0518, 0.0406, 0.0410, 0.0401, 0.0360], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:21:05,425 INFO [train.py:935] (1/4) Epoch 17, validation: loss=0.1774, simple_loss=0.2777, pruned_loss=0.03857, over 944034.00 frames. +2023-02-06 20:21:05,426 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 20:21:12,597 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135339.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:21:16,159 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0229, 2.6177, 3.7084, 2.0187, 1.8684, 3.6746, 0.8088, 2.0747], + device='cuda:1'), covar=tensor([0.1423, 0.1134, 0.0193, 0.1990, 0.3091, 0.0400, 0.2564, 0.1879], + device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0184, 0.0115, 0.0216, 0.0260, 0.0123, 0.0165, 0.0180], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 20:21:36,688 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.411e+02 3.026e+02 3.580e+02 8.983e+02, threshold=6.051e+02, percent-clipped=2.0 +2023-02-06 20:21:40,883 INFO [train.py:901] (1/4) Epoch 17, batch 6050, loss[loss=0.1752, simple_loss=0.2559, pruned_loss=0.04725, over 7635.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.295, pruned_loss=0.06824, over 1615266.16 frames. ], batch size: 19, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:21:41,749 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6567, 4.5993, 4.1167, 2.1774, 4.0407, 4.1948, 4.2320, 3.8872], + device='cuda:1'), covar=tensor([0.0716, 0.0507, 0.0966, 0.4694, 0.0868, 0.0931, 0.1246, 0.0860], + device='cuda:1'), in_proj_covar=tensor([0.0501, 0.0412, 0.0416, 0.0517, 0.0406, 0.0410, 0.0400, 0.0359], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:22:16,312 INFO [train.py:901] (1/4) Epoch 17, batch 6100, loss[loss=0.2144, simple_loss=0.3053, pruned_loss=0.06172, over 8340.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2948, pruned_loss=0.06827, over 1610202.78 frames. ], batch size: 24, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:22:47,556 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.459e+02 2.890e+02 3.783e+02 6.848e+02, threshold=5.780e+02, percent-clipped=3.0 +2023-02-06 20:22:49,626 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 20:22:52,257 INFO [train.py:901] (1/4) Epoch 17, batch 6150, loss[loss=0.1891, simple_loss=0.2892, pruned_loss=0.04451, over 8462.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2949, pruned_loss=0.06797, over 1610211.28 frames. ], batch size: 25, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:23:20,219 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-06 20:23:26,589 INFO [train.py:901] (1/4) Epoch 17, batch 6200, loss[loss=0.2202, simple_loss=0.3065, pruned_loss=0.06692, over 8296.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2939, pruned_loss=0.06701, over 1610747.59 frames. ], batch size: 23, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:23:29,343 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=135533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:23:57,602 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.470e+02 3.035e+02 3.550e+02 6.137e+02, threshold=6.070e+02, percent-clipped=1.0 +2023-02-06 20:24:01,730 INFO [train.py:901] (1/4) Epoch 17, batch 6250, loss[loss=0.1814, simple_loss=0.2732, pruned_loss=0.04481, over 8331.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2933, pruned_loss=0.06651, over 1610900.86 frames. ], batch size: 25, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:24:13,152 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135595.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:24:29,699 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 20:24:30,230 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:24:36,952 INFO [train.py:901] (1/4) Epoch 17, batch 6300, loss[loss=0.2042, simple_loss=0.2861, pruned_loss=0.06115, over 8348.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2932, pruned_loss=0.06652, over 1612931.06 frames. ], batch size: 26, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:24:49,741 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135648.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:24:49,777 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135648.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:25:07,390 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.756e+02 2.700e+02 3.426e+02 4.477e+02 8.691e+02, threshold=6.853e+02, percent-clipped=8.0 +2023-02-06 20:25:11,447 INFO [train.py:901] (1/4) Epoch 17, batch 6350, loss[loss=0.1906, simple_loss=0.267, pruned_loss=0.05712, over 7815.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.295, pruned_loss=0.06693, over 1616215.36 frames. ], batch size: 20, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:25:46,617 INFO [train.py:901] (1/4) Epoch 17, batch 6400, loss[loss=0.1539, simple_loss=0.2338, pruned_loss=0.037, over 7702.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2946, pruned_loss=0.06693, over 1615084.56 frames. ], batch size: 18, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:26:16,682 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.216e+02 2.648e+02 3.143e+02 6.334e+02, threshold=5.295e+02, percent-clipped=0.0 +2023-02-06 20:26:20,491 INFO [train.py:901] (1/4) Epoch 17, batch 6450, loss[loss=0.2557, simple_loss=0.3226, pruned_loss=0.09435, over 8070.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2952, pruned_loss=0.0675, over 1617448.73 frames. ], batch size: 21, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:26:39,795 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2069, 2.2582, 1.9419, 2.9915, 1.4152, 1.6966, 1.8433, 2.3390], + device='cuda:1'), covar=tensor([0.0647, 0.0819, 0.0923, 0.0307, 0.1127, 0.1326, 0.1052, 0.0743], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0198, 0.0248, 0.0211, 0.0209, 0.0248, 0.0256, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 20:26:56,410 INFO [train.py:901] (1/4) Epoch 17, batch 6500, loss[loss=0.2083, simple_loss=0.2974, pruned_loss=0.0596, over 8290.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2957, pruned_loss=0.06771, over 1619935.52 frames. ], batch size: 23, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:27:27,344 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.443e+02 3.095e+02 4.367e+02 8.897e+02, threshold=6.190e+02, percent-clipped=12.0 +2023-02-06 20:27:31,529 INFO [train.py:901] (1/4) Epoch 17, batch 6550, loss[loss=0.2157, simple_loss=0.3032, pruned_loss=0.0641, over 8300.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2944, pruned_loss=0.06693, over 1623015.12 frames. ], batch size: 23, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:27:44,297 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5644, 1.9388, 2.1675, 1.3564, 2.1566, 1.4207, 0.7022, 1.8254], + device='cuda:1'), covar=tensor([0.0572, 0.0301, 0.0231, 0.0514, 0.0379, 0.0769, 0.0693, 0.0290], + device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0372, 0.0320, 0.0425, 0.0353, 0.0512, 0.0374, 0.0395], + device='cuda:1'), out_proj_covar=tensor([1.1723e-04, 9.8929e-05, 8.4846e-05, 1.1351e-04, 9.4554e-05, 1.4739e-04, + 1.0180e-04, 1.0568e-04], device='cuda:1') +2023-02-06 20:27:48,292 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135904.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:27:56,455 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 20:28:06,535 INFO [train.py:901] (1/4) Epoch 17, batch 6600, loss[loss=0.2296, simple_loss=0.3201, pruned_loss=0.0695, over 8108.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2947, pruned_loss=0.06721, over 1618522.41 frames. ], batch size: 23, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:28:06,744 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135929.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:28:16,345 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 20:28:36,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.554e+02 2.943e+02 3.634e+02 1.271e+03, threshold=5.887e+02, percent-clipped=2.0 +2023-02-06 20:28:40,558 INFO [train.py:901] (1/4) Epoch 17, batch 6650, loss[loss=0.2218, simple_loss=0.3115, pruned_loss=0.06605, over 8197.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2931, pruned_loss=0.06631, over 1614286.10 frames. ], batch size: 23, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:28:49,898 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=135992.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:28:56,440 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136000.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:29:04,529 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136012.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:29:12,172 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.17 vs. limit=5.0 +2023-02-06 20:29:16,446 INFO [train.py:901] (1/4) Epoch 17, batch 6700, loss[loss=0.1659, simple_loss=0.2438, pruned_loss=0.04394, over 7810.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2933, pruned_loss=0.06691, over 1614029.10 frames. ], batch size: 19, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:29:45,902 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1051, 1.5126, 1.7576, 1.5120, 0.9425, 1.5590, 1.7323, 1.4275], + device='cuda:1'), covar=tensor([0.0490, 0.1243, 0.1630, 0.1372, 0.0619, 0.1457, 0.0689, 0.0668], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0158, 0.0100, 0.0162, 0.0114, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 20:29:47,782 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.437e+02 3.090e+02 3.837e+02 8.578e+02, threshold=6.181e+02, percent-clipped=4.0 +2023-02-06 20:29:50,646 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7024, 2.1775, 4.2701, 1.5637, 3.0179, 2.3499, 1.7541, 3.0116], + device='cuda:1'), covar=tensor([0.1864, 0.2581, 0.0579, 0.4184, 0.1739, 0.2927, 0.2217, 0.2210], + device='cuda:1'), in_proj_covar=tensor([0.0507, 0.0568, 0.0542, 0.0612, 0.0632, 0.0572, 0.0507, 0.0618], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:29:51,768 INFO [train.py:901] (1/4) Epoch 17, batch 6750, loss[loss=0.2129, simple_loss=0.2971, pruned_loss=0.06431, over 8242.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2947, pruned_loss=0.06759, over 1614624.45 frames. ], batch size: 24, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:30:09,237 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 20:30:11,742 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:30:26,346 INFO [train.py:901] (1/4) Epoch 17, batch 6800, loss[loss=0.1902, simple_loss=0.2745, pruned_loss=0.05295, over 8107.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2934, pruned_loss=0.0668, over 1609137.52 frames. ], batch size: 23, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:30:35,099 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 20:30:48,375 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136160.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:30:57,860 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.691e+02 3.204e+02 3.737e+02 8.793e+02, threshold=6.409e+02, percent-clipped=5.0 +2023-02-06 20:31:01,823 INFO [train.py:901] (1/4) Epoch 17, batch 6850, loss[loss=0.2126, simple_loss=0.2844, pruned_loss=0.07044, over 7412.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2954, pruned_loss=0.06785, over 1609889.55 frames. ], batch size: 17, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:31:22,738 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 20:31:37,004 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 20:31:37,309 INFO [train.py:901] (1/4) Epoch 17, batch 6900, loss[loss=0.1968, simple_loss=0.2731, pruned_loss=0.06022, over 7650.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.294, pruned_loss=0.06714, over 1608282.30 frames. ], batch size: 19, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:32:08,492 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.868e+02 2.541e+02 3.415e+02 4.318e+02 7.722e+02, threshold=6.831e+02, percent-clipped=4.0 +2023-02-06 20:32:12,495 INFO [train.py:901] (1/4) Epoch 17, batch 6950, loss[loss=0.2356, simple_loss=0.3211, pruned_loss=0.07503, over 8242.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.294, pruned_loss=0.06722, over 1606274.72 frames. ], batch size: 24, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:32:32,958 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 20:32:48,141 INFO [train.py:901] (1/4) Epoch 17, batch 7000, loss[loss=0.2237, simple_loss=0.304, pruned_loss=0.07167, over 8549.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.293, pruned_loss=0.06642, over 1608230.13 frames. ], batch size: 49, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:32:54,375 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1220, 1.7272, 4.4119, 2.0211, 2.4144, 4.9601, 5.0676, 4.3119], + device='cuda:1'), covar=tensor([0.1367, 0.1875, 0.0280, 0.2041, 0.1368, 0.0198, 0.0469, 0.0575], + device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0311, 0.0277, 0.0306, 0.0296, 0.0254, 0.0392, 0.0300], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 20:32:58,256 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=136344.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:33:04,331 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136353.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:33:06,277 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=136356.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:33:11,084 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136363.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:33:18,224 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.499e+02 2.956e+02 3.583e+02 7.307e+02, threshold=5.911e+02, percent-clipped=2.0 +2023-02-06 20:33:22,380 INFO [train.py:901] (1/4) Epoch 17, batch 7050, loss[loss=0.2055, simple_loss=0.2887, pruned_loss=0.06113, over 8499.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2934, pruned_loss=0.0669, over 1609073.64 frames. ], batch size: 26, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:33:29,395 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:33:41,804 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0270, 1.6347, 1.3509, 1.5042, 1.3439, 1.1478, 1.2296, 1.2205], + device='cuda:1'), covar=tensor([0.1092, 0.0439, 0.1190, 0.0543, 0.0703, 0.1513, 0.0884, 0.0830], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0229, 0.0322, 0.0297, 0.0293, 0.0327, 0.0338, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 20:33:57,947 INFO [train.py:901] (1/4) Epoch 17, batch 7100, loss[loss=0.2332, simple_loss=0.3165, pruned_loss=0.07498, over 8604.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2941, pruned_loss=0.06705, over 1615102.69 frames. ], batch size: 34, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:34:18,673 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136459.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:34:26,516 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136471.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:34:27,626 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.503e+02 2.917e+02 3.905e+02 1.004e+03, threshold=5.834e+02, percent-clipped=4.0 +2023-02-06 20:34:31,733 INFO [train.py:901] (1/4) Epoch 17, batch 7150, loss[loss=0.1894, simple_loss=0.2709, pruned_loss=0.05393, over 8292.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2945, pruned_loss=0.06751, over 1618008.25 frames. ], batch size: 23, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:34:50,070 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=136504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:35:07,627 INFO [train.py:901] (1/4) Epoch 17, batch 7200, loss[loss=0.2045, simple_loss=0.2879, pruned_loss=0.06053, over 8489.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2934, pruned_loss=0.06689, over 1616269.88 frames. ], batch size: 26, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:35:09,675 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.5161, 1.1927, 3.8427, 1.5200, 3.0895, 3.0612, 3.4166, 3.3858], + device='cuda:1'), covar=tensor([0.1151, 0.6292, 0.1071, 0.4730, 0.2032, 0.1636, 0.1086, 0.1106], + device='cuda:1'), in_proj_covar=tensor([0.0570, 0.0615, 0.0648, 0.0586, 0.0667, 0.0565, 0.0567, 0.0629], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 20:35:20,775 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 20:35:21,979 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6615, 2.0007, 2.2051, 1.2997, 2.2683, 1.5669, 0.6500, 1.8865], + device='cuda:1'), covar=tensor([0.0514, 0.0341, 0.0270, 0.0536, 0.0330, 0.0773, 0.0679, 0.0265], + device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0368, 0.0317, 0.0420, 0.0349, 0.0513, 0.0372, 0.0392], + device='cuda:1'), out_proj_covar=tensor([1.1636e-04, 9.7651e-05, 8.3902e-05, 1.1178e-04, 9.3374e-05, 1.4780e-04, + 1.0136e-04, 1.0500e-04], device='cuda:1') +2023-02-06 20:35:37,959 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.498e+02 3.072e+02 3.698e+02 8.742e+02, threshold=6.145e+02, percent-clipped=2.0 +2023-02-06 20:35:42,152 INFO [train.py:901] (1/4) Epoch 17, batch 7250, loss[loss=0.2386, simple_loss=0.3202, pruned_loss=0.07847, over 8291.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2931, pruned_loss=0.06656, over 1617104.14 frames. ], batch size: 23, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:36:11,359 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136619.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:36:17,806 INFO [train.py:901] (1/4) Epoch 17, batch 7300, loss[loss=0.2384, simple_loss=0.3248, pruned_loss=0.07599, over 7972.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2929, pruned_loss=0.06613, over 1617711.31 frames. ], batch size: 21, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:36:40,630 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:36:48,512 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.506e+02 2.969e+02 3.762e+02 7.100e+02, threshold=5.939e+02, percent-clipped=2.0 +2023-02-06 20:36:52,578 INFO [train.py:901] (1/4) Epoch 17, batch 7350, loss[loss=0.1989, simple_loss=0.2884, pruned_loss=0.05468, over 8110.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2937, pruned_loss=0.06696, over 1612648.64 frames. ], batch size: 23, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:37:05,620 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=136697.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:16,518 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 20:37:17,971 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:27,017 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136727.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:28,129 INFO [train.py:901] (1/4) Epoch 17, batch 7400, loss[loss=0.2397, simple_loss=0.3172, pruned_loss=0.08107, over 8644.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2952, pruned_loss=0.06734, over 1616741.73 frames. ], batch size: 39, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:37:35,010 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 20:37:36,563 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136740.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:45,390 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136752.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:59,293 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.354e+02 2.898e+02 3.777e+02 7.037e+02, threshold=5.795e+02, percent-clipped=3.0 +2023-02-06 20:38:03,298 INFO [train.py:901] (1/4) Epoch 17, batch 7450, loss[loss=0.1764, simple_loss=0.2568, pruned_loss=0.04793, over 7801.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2951, pruned_loss=0.06736, over 1619462.18 frames. ], batch size: 20, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:38:16,576 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 20:38:26,058 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136812.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:38:34,161 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8232, 1.2856, 3.9423, 1.3901, 3.5018, 3.2117, 3.5529, 3.4219], + device='cuda:1'), covar=tensor([0.0603, 0.4530, 0.0563, 0.4332, 0.1132, 0.1021, 0.0645, 0.0778], + device='cuda:1'), in_proj_covar=tensor([0.0577, 0.0622, 0.0656, 0.0596, 0.0673, 0.0574, 0.0574, 0.0634], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 20:38:34,170 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:38:37,330 INFO [train.py:901] (1/4) Epoch 17, batch 7500, loss[loss=0.2068, simple_loss=0.2939, pruned_loss=0.05988, over 8558.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2953, pruned_loss=0.06723, over 1625423.13 frames. ], batch size: 34, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:38:41,765 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 20:39:09,456 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.732e+02 2.462e+02 2.866e+02 3.948e+02 7.787e+02, threshold=5.732e+02, percent-clipped=6.0 +2023-02-06 20:39:11,049 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:39:13,443 INFO [train.py:901] (1/4) Epoch 17, batch 7550, loss[loss=0.2161, simple_loss=0.2867, pruned_loss=0.07278, over 8580.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2943, pruned_loss=0.06679, over 1618932.86 frames. ], batch size: 34, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:39:28,636 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136900.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:39:49,028 INFO [train.py:901] (1/4) Epoch 17, batch 7600, loss[loss=0.1994, simple_loss=0.2884, pruned_loss=0.05524, over 8317.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2943, pruned_loss=0.06725, over 1622659.88 frames. ], batch size: 25, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:40:03,457 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 20:40:14,525 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6361, 2.2160, 4.1917, 1.3669, 2.9826, 2.1113, 1.6359, 2.9183], + device='cuda:1'), covar=tensor([0.1806, 0.2545, 0.0676, 0.4379, 0.1744, 0.3111, 0.2216, 0.2271], + device='cuda:1'), in_proj_covar=tensor([0.0509, 0.0571, 0.0545, 0.0614, 0.0635, 0.0576, 0.0509, 0.0619], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:40:21,094 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.516e+02 2.945e+02 3.717e+02 7.457e+02, threshold=5.891e+02, percent-clipped=6.0 +2023-02-06 20:40:25,201 INFO [train.py:901] (1/4) Epoch 17, batch 7650, loss[loss=0.2333, simple_loss=0.3078, pruned_loss=0.07937, over 8246.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2948, pruned_loss=0.06733, over 1624810.86 frames. ], batch size: 24, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:40:27,414 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1257, 1.8460, 2.5450, 2.0541, 2.3999, 2.2008, 1.8803, 1.2008], + device='cuda:1'), covar=tensor([0.5066, 0.4882, 0.1618, 0.3201, 0.2508, 0.2764, 0.1887, 0.5160], + device='cuda:1'), in_proj_covar=tensor([0.0920, 0.0937, 0.0775, 0.0902, 0.0969, 0.0853, 0.0723, 0.0798], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 20:40:38,677 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7059, 1.6420, 2.4760, 1.9166, 2.1608, 1.6909, 1.4553, 0.9347], + device='cuda:1'), covar=tensor([0.6887, 0.5640, 0.1862, 0.3594, 0.2844, 0.4185, 0.2964, 0.5443], + device='cuda:1'), in_proj_covar=tensor([0.0918, 0.0934, 0.0773, 0.0900, 0.0966, 0.0850, 0.0720, 0.0796], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 20:40:43,283 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=137005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:40:58,549 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 20:41:00,048 INFO [train.py:901] (1/4) Epoch 17, batch 7700, loss[loss=0.2173, simple_loss=0.2947, pruned_loss=0.06994, over 8525.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2956, pruned_loss=0.06776, over 1628139.34 frames. ], batch size: 28, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:41:26,488 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 20:41:26,686 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137068.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:41:30,556 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.598e+02 3.111e+02 3.900e+02 8.834e+02, threshold=6.222e+02, percent-clipped=1.0 +2023-02-06 20:41:34,727 INFO [train.py:901] (1/4) Epoch 17, batch 7750, loss[loss=0.206, simple_loss=0.2886, pruned_loss=0.06173, over 8134.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2952, pruned_loss=0.06776, over 1623435.44 frames. ], batch size: 22, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:41:35,627 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7035, 1.6366, 2.2317, 1.5126, 1.0984, 2.3343, 0.3744, 1.3954], + device='cuda:1'), covar=tensor([0.1683, 0.1286, 0.0396, 0.1451, 0.3430, 0.0438, 0.2610, 0.1645], + device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0183, 0.0116, 0.0216, 0.0263, 0.0123, 0.0166, 0.0182], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 20:41:44,965 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137093.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:42:03,588 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:42:09,492 INFO [train.py:901] (1/4) Epoch 17, batch 7800, loss[loss=0.2584, simple_loss=0.3244, pruned_loss=0.09624, over 7125.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2953, pruned_loss=0.06814, over 1620106.09 frames. ], batch size: 71, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:42:36,424 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=137168.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:42:39,575 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.454e+02 2.768e+02 3.488e+02 7.043e+02, threshold=5.537e+02, percent-clipped=4.0 +2023-02-06 20:42:41,431 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-06 20:42:43,620 INFO [train.py:901] (1/4) Epoch 17, batch 7850, loss[loss=0.2951, simple_loss=0.3593, pruned_loss=0.1155, over 7129.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2966, pruned_loss=0.06866, over 1621118.55 frames. ], batch size: 71, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:43:16,605 INFO [train.py:901] (1/4) Epoch 17, batch 7900, loss[loss=0.2369, simple_loss=0.3174, pruned_loss=0.07822, over 8644.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2972, pruned_loss=0.06923, over 1620882.21 frames. ], batch size: 39, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:43:45,783 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.478e+02 3.005e+02 3.961e+02 6.905e+02, threshold=6.010e+02, percent-clipped=7.0 +2023-02-06 20:43:49,876 INFO [train.py:901] (1/4) Epoch 17, batch 7950, loss[loss=0.1919, simple_loss=0.2813, pruned_loss=0.05125, over 8491.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2964, pruned_loss=0.06862, over 1619045.60 frames. ], batch size: 26, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:43:52,850 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:44:23,211 INFO [train.py:901] (1/4) Epoch 17, batch 8000, loss[loss=0.2709, simple_loss=0.3294, pruned_loss=0.1062, over 7143.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2966, pruned_loss=0.06865, over 1614925.63 frames. ], batch size: 71, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:44:52,962 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.583e+02 3.026e+02 3.684e+02 1.341e+03, threshold=6.053e+02, percent-clipped=4.0 +2023-02-06 20:44:55,370 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:44:57,246 INFO [train.py:901] (1/4) Epoch 17, batch 8050, loss[loss=0.2171, simple_loss=0.2994, pruned_loss=0.06733, over 7919.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2953, pruned_loss=0.0687, over 1604278.08 frames. ], batch size: 20, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:45:12,518 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137401.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:45:14,065 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.77 vs. limit=5.0 +2023-02-06 20:45:29,702 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 20:45:34,949 INFO [train.py:901] (1/4) Epoch 18, batch 0, loss[loss=0.2273, simple_loss=0.3044, pruned_loss=0.07512, over 8371.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3044, pruned_loss=0.07512, over 8371.00 frames. ], batch size: 24, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:45:34,949 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 20:45:46,126 INFO [train.py:935] (1/4) Epoch 18, validation: loss=0.1783, simple_loss=0.2784, pruned_loss=0.03907, over 944034.00 frames. +2023-02-06 20:45:46,128 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 20:46:00,871 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 20:46:20,796 INFO [train.py:901] (1/4) Epoch 18, batch 50, loss[loss=0.2046, simple_loss=0.2802, pruned_loss=0.06446, over 7440.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2918, pruned_loss=0.06506, over 363008.00 frames. ], batch size: 17, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:46:28,994 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 2.698e+02 3.585e+02 4.414e+02 8.769e+02, threshold=7.169e+02, percent-clipped=9.0 +2023-02-06 20:46:35,875 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 20:46:42,986 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6738, 2.2123, 4.1828, 1.4629, 3.1904, 2.1811, 1.6836, 3.0387], + device='cuda:1'), covar=tensor([0.1781, 0.2535, 0.0718, 0.4260, 0.1554, 0.3120, 0.2289, 0.2198], + device='cuda:1'), in_proj_covar=tensor([0.0512, 0.0577, 0.0548, 0.0620, 0.0642, 0.0584, 0.0514, 0.0629], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:46:56,057 INFO [train.py:901] (1/4) Epoch 18, batch 100, loss[loss=0.1988, simple_loss=0.2846, pruned_loss=0.05654, over 8323.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2968, pruned_loss=0.0673, over 644923.62 frames. ], batch size: 25, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:46:58,846 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 20:47:16,434 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137539.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:47:31,280 INFO [train.py:901] (1/4) Epoch 18, batch 150, loss[loss=0.2355, simple_loss=0.3136, pruned_loss=0.07868, over 8674.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2967, pruned_loss=0.06776, over 862960.65 frames. ], batch size: 40, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:47:33,470 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137564.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:47:39,694 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.673e+02 2.369e+02 2.797e+02 3.885e+02 6.122e+02, threshold=5.595e+02, percent-clipped=0.0 +2023-02-06 20:48:07,690 INFO [train.py:901] (1/4) Epoch 18, batch 200, loss[loss=0.2169, simple_loss=0.3106, pruned_loss=0.06156, over 8444.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2947, pruned_loss=0.06687, over 1028948.66 frames. ], batch size: 49, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:48:44,087 INFO [train.py:901] (1/4) Epoch 18, batch 250, loss[loss=0.2683, simple_loss=0.3286, pruned_loss=0.104, over 8145.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2972, pruned_loss=0.06808, over 1159308.42 frames. ], batch size: 22, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:48:51,260 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6944, 1.8860, 1.7447, 2.3007, 0.9555, 1.4358, 1.7207, 1.9219], + device='cuda:1'), covar=tensor([0.0849, 0.0765, 0.0990, 0.0450, 0.1079, 0.1353, 0.0785, 0.0660], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0198, 0.0251, 0.0211, 0.0208, 0.0248, 0.0254, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 20:48:52,398 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.467e+02 3.008e+02 3.586e+02 6.135e+02, threshold=6.015e+02, percent-clipped=1.0 +2023-02-06 20:48:55,926 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 20:49:03,701 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 20:49:19,870 INFO [train.py:901] (1/4) Epoch 18, batch 300, loss[loss=0.1698, simple_loss=0.2543, pruned_loss=0.04269, over 7658.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2947, pruned_loss=0.06713, over 1254296.97 frames. ], batch size: 19, lr: 4.28e-03, grad_scale: 8.0 +2023-02-06 20:49:55,786 INFO [train.py:901] (1/4) Epoch 18, batch 350, loss[loss=0.249, simple_loss=0.3222, pruned_loss=0.08789, over 8343.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2966, pruned_loss=0.06837, over 1338313.04 frames. ], batch size: 24, lr: 4.28e-03, grad_scale: 8.0 +2023-02-06 20:50:03,096 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2717, 2.2775, 2.0549, 2.8769, 1.3036, 1.7653, 1.9755, 2.4014], + device='cuda:1'), covar=tensor([0.0631, 0.0745, 0.0908, 0.0313, 0.1088, 0.1220, 0.0946, 0.0730], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0198, 0.0251, 0.0211, 0.0208, 0.0249, 0.0255, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 20:50:05,697 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.555e+02 3.034e+02 3.752e+02 7.695e+02, threshold=6.069e+02, percent-clipped=3.0 +2023-02-06 20:50:14,084 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2225, 1.3333, 3.3673, 1.0406, 2.9969, 2.8924, 3.1090, 3.0204], + device='cuda:1'), covar=tensor([0.0817, 0.4150, 0.0784, 0.4104, 0.1342, 0.0991, 0.0777, 0.0891], + device='cuda:1'), in_proj_covar=tensor([0.0578, 0.0622, 0.0660, 0.0594, 0.0673, 0.0576, 0.0574, 0.0639], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 20:50:32,304 INFO [train.py:901] (1/4) Epoch 18, batch 400, loss[loss=0.1808, simple_loss=0.2647, pruned_loss=0.04849, over 7650.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2948, pruned_loss=0.06708, over 1399056.43 frames. ], batch size: 19, lr: 4.28e-03, grad_scale: 8.0 +2023-02-06 20:50:33,937 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6764, 1.6546, 2.0978, 1.4926, 1.1606, 2.0888, 0.3014, 1.2754], + device='cuda:1'), covar=tensor([0.1798, 0.1545, 0.0388, 0.1278, 0.3302, 0.0475, 0.2487, 0.1496], + device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0184, 0.0118, 0.0219, 0.0263, 0.0124, 0.0166, 0.0182], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 20:51:08,173 INFO [train.py:901] (1/4) Epoch 18, batch 450, loss[loss=0.2058, simple_loss=0.2796, pruned_loss=0.06601, over 7634.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2937, pruned_loss=0.06694, over 1446314.93 frames. ], batch size: 19, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:51:16,921 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.504e+02 3.016e+02 3.557e+02 6.367e+02, threshold=6.032e+02, percent-clipped=3.0 +2023-02-06 20:51:35,830 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([6.1247, 1.8456, 6.2461, 2.3946, 5.7769, 5.3746, 5.7984, 5.7370], + device='cuda:1'), covar=tensor([0.0420, 0.4195, 0.0274, 0.3267, 0.0814, 0.0792, 0.0458, 0.0421], + device='cuda:1'), in_proj_covar=tensor([0.0576, 0.0617, 0.0656, 0.0591, 0.0670, 0.0574, 0.0570, 0.0634], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 20:51:43,037 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137910.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:51:43,596 INFO [train.py:901] (1/4) Epoch 18, batch 500, loss[loss=0.1894, simple_loss=0.2731, pruned_loss=0.05282, over 7806.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.293, pruned_loss=0.0664, over 1485010.29 frames. ], batch size: 20, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:51:45,174 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137913.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:51:50,427 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 20:52:20,562 INFO [train.py:901] (1/4) Epoch 18, batch 550, loss[loss=0.3221, simple_loss=0.3803, pruned_loss=0.1319, over 8451.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.294, pruned_loss=0.06681, over 1513438.08 frames. ], batch size: 27, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:52:29,489 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.606e+02 3.197e+02 3.974e+02 7.545e+02, threshold=6.394e+02, percent-clipped=3.0 +2023-02-06 20:52:35,967 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 20:52:56,986 INFO [train.py:901] (1/4) Epoch 18, batch 600, loss[loss=0.2236, simple_loss=0.3074, pruned_loss=0.0699, over 7806.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2934, pruned_loss=0.06662, over 1534419.10 frames. ], batch size: 20, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:53:11,901 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 20:53:32,792 INFO [train.py:901] (1/4) Epoch 18, batch 650, loss[loss=0.1936, simple_loss=0.278, pruned_loss=0.05465, over 8148.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2917, pruned_loss=0.06551, over 1549881.53 frames. ], batch size: 22, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:53:43,407 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.461e+02 2.865e+02 3.365e+02 7.739e+02, threshold=5.729e+02, percent-clipped=1.0 +2023-02-06 20:54:00,541 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 20:54:09,457 INFO [train.py:901] (1/4) Epoch 18, batch 700, loss[loss=0.236, simple_loss=0.3059, pruned_loss=0.08307, over 8077.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2923, pruned_loss=0.06571, over 1564071.48 frames. ], batch size: 21, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:54:29,538 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 20:54:30,792 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7066, 1.7049, 2.0172, 1.4595, 1.2784, 2.1493, 0.2877, 1.3200], + device='cuda:1'), covar=tensor([0.1922, 0.1314, 0.0498, 0.1299, 0.3019, 0.0457, 0.2321, 0.1586], + device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0184, 0.0117, 0.0217, 0.0262, 0.0123, 0.0165, 0.0182], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 20:54:44,136 INFO [train.py:901] (1/4) Epoch 18, batch 750, loss[loss=0.2083, simple_loss=0.297, pruned_loss=0.05987, over 8109.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2925, pruned_loss=0.06556, over 1578503.89 frames. ], batch size: 23, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:54:53,226 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.332e+02 3.041e+02 3.730e+02 6.216e+02, threshold=6.081e+02, percent-clipped=3.0 +2023-02-06 20:54:58,171 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 20:55:08,082 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 20:55:19,816 INFO [train.py:901] (1/4) Epoch 18, batch 800, loss[loss=0.2109, simple_loss=0.296, pruned_loss=0.06293, over 8128.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2923, pruned_loss=0.06535, over 1586081.80 frames. ], batch size: 22, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:55:49,564 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=138254.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:55:51,610 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=138257.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:55:54,216 INFO [train.py:901] (1/4) Epoch 18, batch 850, loss[loss=0.227, simple_loss=0.3077, pruned_loss=0.07316, over 8448.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2931, pruned_loss=0.06652, over 1591201.25 frames. ], batch size: 27, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:56:03,044 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.308e+02 2.906e+02 3.562e+02 8.427e+02, threshold=5.812e+02, percent-clipped=4.0 +2023-02-06 20:56:13,564 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:56:30,845 INFO [train.py:901] (1/4) Epoch 18, batch 900, loss[loss=0.2418, simple_loss=0.3133, pruned_loss=0.08518, over 7660.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.294, pruned_loss=0.06649, over 1598393.20 frames. ], batch size: 19, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:56:35,140 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1105, 1.7289, 2.1578, 1.7869, 1.3351, 1.7187, 2.3923, 2.4754], + device='cuda:1'), covar=tensor([0.0394, 0.1184, 0.1534, 0.1293, 0.0518, 0.1386, 0.0557, 0.0496], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0190, 0.0158, 0.0100, 0.0161, 0.0114, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 20:56:50,623 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138340.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:57:05,377 INFO [train.py:901] (1/4) Epoch 18, batch 950, loss[loss=0.174, simple_loss=0.2545, pruned_loss=0.04675, over 7407.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2929, pruned_loss=0.06591, over 1600596.86 frames. ], batch size: 17, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:57:10,959 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138369.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:57:13,027 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138372.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:57:14,180 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.844e+02 2.531e+02 3.020e+02 3.937e+02 8.991e+02, threshold=6.039e+02, percent-clipped=7.0 +2023-02-06 20:57:14,402 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8716, 2.1046, 1.8513, 2.7568, 1.2473, 1.6140, 1.9767, 2.3395], + device='cuda:1'), covar=tensor([0.0751, 0.0748, 0.0862, 0.0314, 0.1044, 0.1254, 0.0840, 0.0667], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0200, 0.0251, 0.0212, 0.0207, 0.0249, 0.0253, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 20:57:29,241 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 20:57:40,381 INFO [train.py:901] (1/4) Epoch 18, batch 1000, loss[loss=0.2275, simple_loss=0.3242, pruned_loss=0.06536, over 8454.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2936, pruned_loss=0.06663, over 1604917.76 frames. ], batch size: 27, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:57:56,198 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5944, 1.7965, 2.6833, 1.4096, 1.8338, 1.9201, 1.6272, 1.7620], + device='cuda:1'), covar=tensor([0.1930, 0.2695, 0.0930, 0.4278, 0.1980, 0.3145, 0.2194, 0.2348], + device='cuda:1'), in_proj_covar=tensor([0.0511, 0.0574, 0.0544, 0.0618, 0.0636, 0.0577, 0.0510, 0.0624], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:58:05,490 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 20:58:16,697 INFO [train.py:901] (1/4) Epoch 18, batch 1050, loss[loss=0.2111, simple_loss=0.296, pruned_loss=0.06307, over 8134.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2923, pruned_loss=0.06589, over 1611140.96 frames. ], batch size: 22, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:58:18,829 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 20:58:25,548 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.682e+02 2.454e+02 3.228e+02 4.133e+02 8.765e+02, threshold=6.456e+02, percent-clipped=4.0 +2023-02-06 20:58:51,059 INFO [train.py:901] (1/4) Epoch 18, batch 1100, loss[loss=0.2411, simple_loss=0.3224, pruned_loss=0.07992, over 8498.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2923, pruned_loss=0.06562, over 1613509.51 frames. ], batch size: 28, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 20:58:56,193 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2339, 3.1474, 2.9574, 1.7020, 2.8437, 2.8779, 2.8640, 2.7454], + device='cuda:1'), covar=tensor([0.1221, 0.0898, 0.1345, 0.4465, 0.1104, 0.1369, 0.1724, 0.1083], + device='cuda:1'), in_proj_covar=tensor([0.0507, 0.0419, 0.0419, 0.0517, 0.0409, 0.0422, 0.0407, 0.0365], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 20:59:26,911 INFO [train.py:901] (1/4) Epoch 18, batch 1150, loss[loss=0.1903, simple_loss=0.2631, pruned_loss=0.05877, over 7807.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2923, pruned_loss=0.06563, over 1613804.06 frames. ], batch size: 20, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 20:59:29,609 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 20:59:35,880 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 2.366e+02 2.909e+02 3.553e+02 5.350e+02, threshold=5.817e+02, percent-clipped=0.0 +2023-02-06 20:59:44,953 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5003, 1.8994, 2.9489, 1.2515, 2.1482, 1.7246, 1.5903, 2.0681], + device='cuda:1'), covar=tensor([0.2161, 0.2716, 0.0920, 0.4962, 0.2110, 0.3697, 0.2472, 0.2691], + device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0577, 0.0550, 0.0625, 0.0642, 0.0583, 0.0515, 0.0630], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:00:02,034 INFO [train.py:901] (1/4) Epoch 18, batch 1200, loss[loss=0.1843, simple_loss=0.2631, pruned_loss=0.05272, over 7652.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2915, pruned_loss=0.06563, over 1611770.62 frames. ], batch size: 19, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:00:04,277 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2322, 1.2350, 3.3744, 1.0811, 2.9485, 2.8368, 3.0776, 2.9875], + device='cuda:1'), covar=tensor([0.0905, 0.4161, 0.0762, 0.4109, 0.1468, 0.1022, 0.0795, 0.0932], + device='cuda:1'), in_proj_covar=tensor([0.0582, 0.0618, 0.0662, 0.0591, 0.0674, 0.0577, 0.0569, 0.0642], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:00:11,841 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138625.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:00:13,947 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138628.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:00:16,608 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=138632.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:00:29,704 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138650.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:00:31,811 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138653.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:00:37,163 INFO [train.py:901] (1/4) Epoch 18, batch 1250, loss[loss=0.2113, simple_loss=0.2864, pruned_loss=0.06811, over 8086.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2935, pruned_loss=0.06628, over 1617358.92 frames. ], batch size: 21, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:00:46,118 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4074, 4.3864, 4.0423, 1.9004, 3.9950, 4.0225, 4.0080, 3.8372], + device='cuda:1'), covar=tensor([0.0785, 0.0569, 0.1037, 0.4669, 0.0783, 0.0967, 0.1307, 0.0683], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0420, 0.0421, 0.0521, 0.0412, 0.0424, 0.0409, 0.0367], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:00:47,267 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.310e+02 2.834e+02 3.613e+02 5.274e+02, threshold=5.668e+02, percent-clipped=0.0 +2023-02-06 21:00:54,239 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=138684.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:04,836 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138699.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:08,259 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138704.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:12,852 INFO [train.py:901] (1/4) Epoch 18, batch 1300, loss[loss=0.1673, simple_loss=0.2453, pruned_loss=0.04463, over 7807.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2933, pruned_loss=0.06583, over 1621625.97 frames. ], batch size: 20, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:01:15,828 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:34,961 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4094, 1.6681, 2.7252, 1.2614, 1.9421, 1.7713, 1.4283, 1.9075], + device='cuda:1'), covar=tensor([0.1819, 0.2327, 0.0859, 0.4357, 0.1778, 0.3156, 0.2275, 0.2107], + device='cuda:1'), in_proj_covar=tensor([0.0514, 0.0577, 0.0549, 0.0624, 0.0643, 0.0583, 0.0513, 0.0630], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:01:37,544 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:38,939 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8228, 1.9794, 1.7562, 2.5818, 1.1232, 1.5067, 1.7309, 2.1446], + device='cuda:1'), covar=tensor([0.0785, 0.0849, 0.0964, 0.0380, 0.1138, 0.1413, 0.0973, 0.0771], + device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0202, 0.0254, 0.0214, 0.0210, 0.0251, 0.0256, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 21:01:47,850 INFO [train.py:901] (1/4) Epoch 18, batch 1350, loss[loss=0.1948, simple_loss=0.282, pruned_loss=0.05384, over 8066.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2928, pruned_loss=0.06534, over 1625857.99 frames. ], batch size: 21, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:01:56,591 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.404e+02 2.906e+02 3.545e+02 6.613e+02, threshold=5.812e+02, percent-clipped=4.0 +2023-02-06 21:02:15,438 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138799.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:02:23,342 INFO [train.py:901] (1/4) Epoch 18, batch 1400, loss[loss=0.238, simple_loss=0.3013, pruned_loss=0.08738, over 7800.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2942, pruned_loss=0.06654, over 1622195.42 frames. ], batch size: 19, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:02:43,933 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.16 vs. limit=5.0 +2023-02-06 21:02:57,619 INFO [train.py:901] (1/4) Epoch 18, batch 1450, loss[loss=0.2105, simple_loss=0.3032, pruned_loss=0.0589, over 8450.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2929, pruned_loss=0.06591, over 1619762.10 frames. ], batch size: 27, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:03:06,388 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 2.491e+02 3.050e+02 4.246e+02 7.467e+02, threshold=6.100e+02, percent-clipped=3.0 +2023-02-06 21:03:07,085 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 21:03:33,396 INFO [train.py:901] (1/4) Epoch 18, batch 1500, loss[loss=0.192, simple_loss=0.2719, pruned_loss=0.05604, over 8085.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2938, pruned_loss=0.06603, over 1620799.83 frames. ], batch size: 21, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:03:33,624 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7847, 2.1002, 2.2620, 1.3604, 2.3342, 1.5929, 0.7063, 1.9320], + device='cuda:1'), covar=tensor([0.0563, 0.0314, 0.0232, 0.0544, 0.0343, 0.0787, 0.0786, 0.0305], + device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0373, 0.0321, 0.0428, 0.0358, 0.0514, 0.0379, 0.0400], + device='cuda:1'), out_proj_covar=tensor([1.1840e-04, 9.9095e-05, 8.5226e-05, 1.1405e-04, 9.5537e-05, 1.4746e-04, + 1.0318e-04, 1.0699e-04], device='cuda:1') +2023-02-06 21:03:40,046 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4857, 2.0078, 3.1487, 1.2282, 2.3266, 1.8155, 1.5908, 2.2617], + device='cuda:1'), covar=tensor([0.2104, 0.2380, 0.0842, 0.4784, 0.1921, 0.3311, 0.2340, 0.2448], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0577, 0.0549, 0.0622, 0.0643, 0.0581, 0.0512, 0.0630], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:03:40,186 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 21:03:44,796 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:04:01,170 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2666, 1.3226, 1.5675, 1.2543, 0.7400, 1.3568, 1.2062, 1.0473], + device='cuda:1'), covar=tensor([0.0542, 0.1256, 0.1592, 0.1441, 0.0553, 0.1482, 0.0697, 0.0701], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0157, 0.0100, 0.0161, 0.0114, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 21:04:07,342 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9989, 2.6740, 2.1405, 2.3283, 2.3287, 2.0455, 2.1873, 2.4054], + device='cuda:1'), covar=tensor([0.1075, 0.0323, 0.0879, 0.0573, 0.0585, 0.1049, 0.0746, 0.0809], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0236, 0.0328, 0.0303, 0.0298, 0.0330, 0.0342, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 21:04:08,396 INFO [train.py:901] (1/4) Epoch 18, batch 1550, loss[loss=0.2184, simple_loss=0.2837, pruned_loss=0.07649, over 7784.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2945, pruned_loss=0.06693, over 1614682.59 frames. ], batch size: 19, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:04:17,342 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.725e+02 2.366e+02 2.933e+02 3.736e+02 6.367e+02, threshold=5.865e+02, percent-clipped=3.0 +2023-02-06 21:04:38,172 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139003.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:04:43,436 INFO [train.py:901] (1/4) Epoch 18, batch 1600, loss[loss=0.1732, simple_loss=0.2592, pruned_loss=0.04359, over 8242.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2944, pruned_loss=0.06658, over 1616497.05 frames. ], batch size: 22, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:04:56,940 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139028.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:07,021 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:10,356 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139048.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:15,374 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0244, 2.4103, 3.5929, 1.9825, 1.8614, 3.5356, 0.6786, 2.1732], + device='cuda:1'), covar=tensor([0.1484, 0.1407, 0.0240, 0.1881, 0.3048, 0.0424, 0.2480, 0.1593], + device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0188, 0.0118, 0.0217, 0.0264, 0.0126, 0.0164, 0.0184], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 21:05:15,387 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139055.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:18,035 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:20,007 INFO [train.py:901] (1/4) Epoch 18, batch 1650, loss[loss=0.1954, simple_loss=0.2696, pruned_loss=0.06053, over 7275.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2932, pruned_loss=0.06593, over 1614506.71 frames. ], batch size: 16, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:05:28,837 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.409e+02 2.375e+02 2.907e+02 3.508e+02 7.626e+02, threshold=5.813e+02, percent-clipped=3.0 +2023-02-06 21:05:33,239 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:41,376 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-06 21:05:42,208 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-06 21:05:54,415 INFO [train.py:901] (1/4) Epoch 18, batch 1700, loss[loss=0.186, simple_loss=0.2619, pruned_loss=0.05506, over 8091.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2917, pruned_loss=0.06512, over 1613547.47 frames. ], batch size: 21, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:06:02,371 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0236, 1.9034, 3.2985, 1.6437, 2.4358, 3.6490, 3.6179, 3.2045], + device='cuda:1'), covar=tensor([0.1070, 0.1558, 0.0373, 0.1930, 0.1176, 0.0203, 0.0565, 0.0486], + device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0318, 0.0279, 0.0310, 0.0301, 0.0257, 0.0399, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 21:06:28,686 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139158.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:06:29,301 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1976, 4.1648, 3.7314, 1.8476, 3.6732, 3.7571, 3.8454, 3.4514], + device='cuda:1'), covar=tensor([0.0742, 0.0527, 0.1064, 0.4764, 0.0923, 0.0897, 0.1210, 0.0848], + device='cuda:1'), in_proj_covar=tensor([0.0505, 0.0414, 0.0417, 0.0521, 0.0412, 0.0423, 0.0408, 0.0364], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:06:30,624 INFO [train.py:901] (1/4) Epoch 18, batch 1750, loss[loss=0.1828, simple_loss=0.2612, pruned_loss=0.05219, over 7794.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2919, pruned_loss=0.06548, over 1615111.44 frames. ], batch size: 20, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:06:32,186 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139163.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:06:39,595 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.310e+02 2.913e+02 3.912e+02 7.750e+02, threshold=5.826e+02, percent-clipped=6.0 +2023-02-06 21:06:39,810 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139174.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:07:05,729 INFO [train.py:901] (1/4) Epoch 18, batch 1800, loss[loss=0.208, simple_loss=0.3043, pruned_loss=0.05583, over 8333.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2918, pruned_loss=0.06538, over 1614522.44 frames. ], batch size: 25, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:07:37,621 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139256.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:07:40,849 INFO [train.py:901] (1/4) Epoch 18, batch 1850, loss[loss=0.2132, simple_loss=0.2816, pruned_loss=0.07235, over 7644.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2919, pruned_loss=0.06539, over 1614150.85 frames. ], batch size: 19, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:07:49,485 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139271.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:07:49,612 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6945, 1.6586, 2.2169, 1.4416, 1.2312, 2.2730, 0.2929, 1.3461], + device='cuda:1'), covar=tensor([0.1749, 0.1313, 0.0342, 0.1477, 0.3201, 0.0409, 0.2700, 0.1512], + device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0187, 0.0117, 0.0216, 0.0261, 0.0125, 0.0164, 0.0183], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 21:07:51,404 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.274e+02 2.776e+02 3.369e+02 8.658e+02, threshold=5.552e+02, percent-clipped=2.0 +2023-02-06 21:08:05,265 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139294.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:08:14,572 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139307.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:08:17,142 INFO [train.py:901] (1/4) Epoch 18, batch 1900, loss[loss=0.2174, simple_loss=0.3076, pruned_loss=0.06365, over 8464.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2931, pruned_loss=0.06607, over 1615083.69 frames. ], batch size: 27, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:08:29,929 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-06 21:08:52,414 INFO [train.py:901] (1/4) Epoch 18, batch 1950, loss[loss=0.2083, simple_loss=0.2946, pruned_loss=0.06098, over 8239.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2942, pruned_loss=0.06635, over 1618473.05 frames. ], batch size: 22, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:08:55,254 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 21:09:01,254 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 2.421e+02 2.964e+02 3.877e+02 7.962e+02, threshold=5.927e+02, percent-clipped=5.0 +2023-02-06 21:09:08,113 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 21:09:11,170 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139386.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:28,238 INFO [train.py:901] (1/4) Epoch 18, batch 2000, loss[loss=0.2336, simple_loss=0.3162, pruned_loss=0.07556, over 8522.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2948, pruned_loss=0.06672, over 1617046.66 frames. ], batch size: 26, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:09:28,243 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 21:09:30,565 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139414.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:33,881 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139419.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:42,124 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139430.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:48,227 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139439.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:49,574 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1076, 2.2850, 1.9684, 2.9241, 1.2883, 1.6494, 1.9405, 2.2873], + device='cuda:1'), covar=tensor([0.0748, 0.0824, 0.0952, 0.0376, 0.1182, 0.1357, 0.0942, 0.0811], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0200, 0.0250, 0.0213, 0.0207, 0.0249, 0.0255, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 21:09:51,650 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139444.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:59,137 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139455.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:10:02,970 INFO [train.py:901] (1/4) Epoch 18, batch 2050, loss[loss=0.2279, simple_loss=0.3138, pruned_loss=0.07098, over 8322.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2933, pruned_loss=0.06617, over 1611838.47 frames. ], batch size: 26, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:10:12,679 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.515e+02 3.080e+02 3.592e+02 7.733e+02, threshold=6.160e+02, percent-clipped=3.0 +2023-02-06 21:10:39,819 INFO [train.py:901] (1/4) Epoch 18, batch 2100, loss[loss=0.2471, simple_loss=0.3086, pruned_loss=0.0928, over 7973.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.294, pruned_loss=0.06638, over 1617304.82 frames. ], batch size: 21, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:11:15,316 INFO [train.py:901] (1/4) Epoch 18, batch 2150, loss[loss=0.1958, simple_loss=0.2892, pruned_loss=0.05116, over 8239.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2943, pruned_loss=0.06643, over 1618343.90 frames. ], batch size: 24, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:11:24,961 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.487e+02 3.024e+02 3.808e+02 9.008e+02, threshold=6.048e+02, percent-clipped=4.0 +2023-02-06 21:11:34,718 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139589.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:11:43,090 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139600.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:11:43,167 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139600.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:11:50,324 INFO [train.py:901] (1/4) Epoch 18, batch 2200, loss[loss=0.2311, simple_loss=0.307, pruned_loss=0.0776, over 8106.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2932, pruned_loss=0.06631, over 1617422.26 frames. ], batch size: 23, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:12:10,495 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139638.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:12:13,211 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139642.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:12:19,372 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139651.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:12:19,555 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5667, 2.0210, 3.2822, 1.2714, 2.3937, 2.0617, 1.6269, 2.4425], + device='cuda:1'), covar=tensor([0.1939, 0.2583, 0.0805, 0.4749, 0.1887, 0.3138, 0.2265, 0.2300], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0577, 0.0554, 0.0622, 0.0640, 0.0577, 0.0513, 0.0627], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:12:26,631 INFO [train.py:901] (1/4) Epoch 18, batch 2250, loss[loss=0.2739, simple_loss=0.3396, pruned_loss=0.1041, over 8131.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2934, pruned_loss=0.06659, over 1617448.88 frames. ], batch size: 22, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:12:31,128 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:12:36,169 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.519e+02 3.270e+02 4.475e+02 8.912e+02, threshold=6.540e+02, percent-clipped=11.0 +2023-02-06 21:13:01,638 INFO [train.py:901] (1/4) Epoch 18, batch 2300, loss[loss=0.246, simple_loss=0.3261, pruned_loss=0.0829, over 8496.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2942, pruned_loss=0.06717, over 1620789.93 frames. ], batch size: 26, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:13:04,658 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:13:32,010 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139753.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:13:37,197 INFO [train.py:901] (1/4) Epoch 18, batch 2350, loss[loss=0.2181, simple_loss=0.3032, pruned_loss=0.06653, over 8653.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2939, pruned_loss=0.0671, over 1618877.92 frames. ], batch size: 34, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:13:40,613 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139766.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:13:44,706 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2396, 1.2789, 3.3678, 1.1185, 2.9545, 2.8089, 3.0461, 2.9663], + device='cuda:1'), covar=tensor([0.0835, 0.4435, 0.0828, 0.4123, 0.1466, 0.1161, 0.0849, 0.0942], + device='cuda:1'), in_proj_covar=tensor([0.0585, 0.0618, 0.0668, 0.0594, 0.0673, 0.0578, 0.0574, 0.0646], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:13:47,225 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.439e+02 2.945e+02 3.859e+02 6.515e+02, threshold=5.891e+02, percent-clipped=0.0 +2023-02-06 21:13:49,999 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9913, 1.5362, 3.5574, 1.5514, 2.3382, 3.9549, 3.9766, 3.4002], + device='cuda:1'), covar=tensor([0.1084, 0.1738, 0.0295, 0.1956, 0.1077, 0.0188, 0.0391, 0.0491], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0314, 0.0274, 0.0306, 0.0295, 0.0254, 0.0394, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 21:13:51,083 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 21:13:55,510 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139787.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:09,008 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:11,589 INFO [train.py:901] (1/4) Epoch 18, batch 2400, loss[loss=0.2219, simple_loss=0.2969, pruned_loss=0.07349, over 7926.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2948, pruned_loss=0.06788, over 1621725.87 frames. ], batch size: 20, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:14:20,250 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139822.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:21,630 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:48,489 INFO [train.py:901] (1/4) Epoch 18, batch 2450, loss[loss=0.2152, simple_loss=0.2884, pruned_loss=0.07099, over 7194.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2946, pruned_loss=0.06806, over 1612498.05 frames. ], batch size: 72, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:14:53,470 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:58,171 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.359e+02 2.854e+02 3.442e+02 8.627e+02, threshold=5.708e+02, percent-clipped=1.0 +2023-02-06 21:15:23,626 INFO [train.py:901] (1/4) Epoch 18, batch 2500, loss[loss=0.1832, simple_loss=0.262, pruned_loss=0.05223, over 7441.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2953, pruned_loss=0.06849, over 1608251.55 frames. ], batch size: 17, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:15:39,669 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139933.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:15:46,142 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.31 vs. limit=5.0 +2023-02-06 21:15:47,235 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:00,244 INFO [train.py:901] (1/4) Epoch 18, batch 2550, loss[loss=0.1925, simple_loss=0.273, pruned_loss=0.056, over 8856.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2942, pruned_loss=0.06782, over 1614217.92 frames. ], batch size: 40, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:16:07,327 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139971.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:09,813 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.329e+02 2.906e+02 3.594e+02 7.294e+02, threshold=5.811e+02, percent-clipped=3.0 +2023-02-06 21:16:25,058 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139996.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:32,756 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6418, 1.9540, 2.0089, 1.2935, 2.1444, 1.4481, 0.6286, 1.9421], + device='cuda:1'), covar=tensor([0.0398, 0.0239, 0.0174, 0.0420, 0.0238, 0.0628, 0.0641, 0.0195], + device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0371, 0.0321, 0.0426, 0.0357, 0.0517, 0.0378, 0.0398], + device='cuda:1'), out_proj_covar=tensor([1.1712e-04, 9.8174e-05, 8.5056e-05, 1.1353e-04, 9.5178e-05, 1.4837e-04, + 1.0278e-04, 1.0658e-04], device='cuda:1') +2023-02-06 21:16:35,505 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140009.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:36,614 INFO [train.py:901] (1/4) Epoch 18, batch 2600, loss[loss=0.2534, simple_loss=0.324, pruned_loss=0.09144, over 8563.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2942, pruned_loss=0.06773, over 1617019.59 frames. ], batch size: 31, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:16:40,965 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0812, 1.3201, 4.3266, 1.5772, 3.7512, 3.5534, 3.8987, 3.7574], + device='cuda:1'), covar=tensor([0.0730, 0.4701, 0.0516, 0.3957, 0.1268, 0.0918, 0.0640, 0.0768], + device='cuda:1'), in_proj_covar=tensor([0.0585, 0.0619, 0.0668, 0.0595, 0.0674, 0.0578, 0.0574, 0.0647], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:16:44,583 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:44,601 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5184, 1.7822, 2.5707, 1.3756, 1.9956, 1.8255, 1.5965, 1.8878], + device='cuda:1'), covar=tensor([0.1617, 0.2168, 0.0775, 0.3928, 0.1580, 0.2767, 0.1964, 0.2049], + device='cuda:1'), in_proj_covar=tensor([0.0511, 0.0575, 0.0552, 0.0620, 0.0637, 0.0578, 0.0513, 0.0626], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:16:52,709 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140034.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:17:01,563 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140047.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:17:03,000 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140048.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:17:05,689 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0682, 2.3633, 1.8822, 2.8098, 1.3726, 1.5413, 1.9932, 2.3027], + device='cuda:1'), covar=tensor([0.0719, 0.0675, 0.0935, 0.0368, 0.1104, 0.1359, 0.0944, 0.0773], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0202, 0.0254, 0.0215, 0.0209, 0.0253, 0.0258, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 21:17:10,537 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:17:11,787 INFO [train.py:901] (1/4) Epoch 18, batch 2650, loss[loss=0.1734, simple_loss=0.269, pruned_loss=0.03896, over 7794.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2957, pruned_loss=0.06804, over 1619413.06 frames. ], batch size: 19, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:17:22,343 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.571e+02 2.973e+02 3.666e+02 6.732e+02, threshold=5.945e+02, percent-clipped=3.0 +2023-02-06 21:17:47,905 INFO [train.py:901] (1/4) Epoch 18, batch 2700, loss[loss=0.1946, simple_loss=0.2837, pruned_loss=0.05278, over 8290.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2958, pruned_loss=0.06837, over 1617639.23 frames. ], batch size: 23, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:17:55,831 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140121.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:00,555 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:02,538 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140131.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:16,408 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140151.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:23,215 INFO [train.py:901] (1/4) Epoch 18, batch 2750, loss[loss=0.2009, simple_loss=0.2799, pruned_loss=0.06093, over 8441.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2954, pruned_loss=0.06817, over 1618762.71 frames. ], batch size: 49, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:18:27,487 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140166.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:28,872 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140168.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:33,791 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.559e+02 2.336e+02 2.919e+02 3.807e+02 8.313e+02, threshold=5.838e+02, percent-clipped=5.0 +2023-02-06 21:19:00,768 INFO [train.py:901] (1/4) Epoch 18, batch 2800, loss[loss=0.1941, simple_loss=0.271, pruned_loss=0.05864, over 8242.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.295, pruned_loss=0.06799, over 1614926.29 frames. ], batch size: 22, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:19:01,489 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140212.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:02,304 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1154, 1.8171, 2.3882, 2.0103, 2.2650, 2.0942, 1.8349, 1.0497], + device='cuda:1'), covar=tensor([0.4626, 0.4335, 0.1618, 0.3178, 0.2113, 0.2795, 0.1888, 0.4823], + device='cuda:1'), in_proj_covar=tensor([0.0929, 0.0949, 0.0785, 0.0914, 0.0981, 0.0867, 0.0732, 0.0812], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:19:25,724 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140246.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:32,578 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1951, 1.4810, 4.3476, 1.7100, 3.8906, 3.6493, 3.9114, 3.7689], + device='cuda:1'), covar=tensor([0.0524, 0.4493, 0.0520, 0.3556, 0.1075, 0.0873, 0.0587, 0.0667], + device='cuda:1'), in_proj_covar=tensor([0.0585, 0.0618, 0.0664, 0.0594, 0.0675, 0.0579, 0.0574, 0.0644], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:19:34,250 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.87 vs. limit=5.0 +2023-02-06 21:19:35,894 INFO [train.py:901] (1/4) Epoch 18, batch 2850, loss[loss=0.2141, simple_loss=0.2797, pruned_loss=0.07423, over 7548.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2944, pruned_loss=0.06728, over 1615277.77 frames. ], batch size: 18, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:19:39,558 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140266.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:45,685 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.859e+02 2.447e+02 2.919e+02 3.574e+02 5.806e+02, threshold=5.838e+02, percent-clipped=0.0 +2023-02-06 21:19:47,302 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140277.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:50,899 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140281.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:52,284 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:53,087 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7514, 2.0760, 2.1917, 1.2637, 2.2486, 1.6164, 0.6628, 1.9176], + device='cuda:1'), covar=tensor([0.0556, 0.0355, 0.0278, 0.0573, 0.0384, 0.0854, 0.0827, 0.0301], + device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0373, 0.0321, 0.0430, 0.0359, 0.0519, 0.0380, 0.0400], + device='cuda:1'), out_proj_covar=tensor([1.1765e-04, 9.8841e-05, 8.4977e-05, 1.1459e-04, 9.5747e-05, 1.4865e-04, + 1.0347e-04, 1.0696e-04], device='cuda:1') +2023-02-06 21:20:07,403 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140304.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:11,940 INFO [train.py:901] (1/4) Epoch 18, batch 2900, loss[loss=0.2154, simple_loss=0.3056, pruned_loss=0.06257, over 8447.00 frames. ], tot_loss[loss=0.215, simple_loss=0.295, pruned_loss=0.06752, over 1613695.60 frames. ], batch size: 27, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:20:14,966 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140315.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:23,758 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:25,183 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140329.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:32,759 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140340.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:44,375 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 21:20:47,994 INFO [train.py:901] (1/4) Epoch 18, batch 2950, loss[loss=0.2073, simple_loss=0.274, pruned_loss=0.07031, over 7534.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.295, pruned_loss=0.06785, over 1613094.09 frames. ], batch size: 18, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:20:57,349 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.671e+02 3.280e+02 4.327e+02 7.160e+02, threshold=6.561e+02, percent-clipped=5.0 +2023-02-06 21:21:02,944 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6249, 1.6616, 4.7915, 1.8315, 4.3617, 3.9773, 4.3154, 4.1908], + device='cuda:1'), covar=tensor([0.0491, 0.4076, 0.0417, 0.3496, 0.0885, 0.0893, 0.0511, 0.0579], + device='cuda:1'), in_proj_covar=tensor([0.0586, 0.0618, 0.0665, 0.0596, 0.0675, 0.0578, 0.0575, 0.0646], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:21:18,951 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140405.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:21:23,798 INFO [train.py:901] (1/4) Epoch 18, batch 3000, loss[loss=0.2414, simple_loss=0.3167, pruned_loss=0.08311, over 6229.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2959, pruned_loss=0.06861, over 1611296.01 frames. ], batch size: 71, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:21:23,798 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 21:21:37,690 INFO [train.py:935] (1/4) Epoch 18, validation: loss=0.1773, simple_loss=0.2774, pruned_loss=0.03861, over 944034.00 frames. +2023-02-06 21:21:37,691 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 21:22:14,080 INFO [train.py:901] (1/4) Epoch 18, batch 3050, loss[loss=0.2428, simple_loss=0.3226, pruned_loss=0.08147, over 8490.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2973, pruned_loss=0.06959, over 1616042.32 frames. ], batch size: 26, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:22:16,898 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140465.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:22:21,665 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:22:24,219 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.663e+02 3.172e+02 4.119e+02 9.916e+02, threshold=6.345e+02, percent-clipped=7.0 +2023-02-06 21:22:40,219 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1798, 1.6210, 4.3153, 1.9997, 2.3727, 4.9154, 4.9758, 4.2393], + device='cuda:1'), covar=tensor([0.1165, 0.1789, 0.0285, 0.1862, 0.1262, 0.0190, 0.0460, 0.0589], + device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0316, 0.0280, 0.0310, 0.0300, 0.0258, 0.0402, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 21:22:42,890 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140502.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:22:46,273 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:22:48,880 INFO [train.py:901] (1/4) Epoch 18, batch 3100, loss[loss=0.1968, simple_loss=0.2974, pruned_loss=0.0481, over 8335.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2982, pruned_loss=0.06972, over 1614084.38 frames. ], batch size: 25, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:22:56,952 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140522.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:00,963 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140527.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:01,023 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140527.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:01,872 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 21:23:07,862 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140537.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:09,255 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140539.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:15,571 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140547.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:25,507 INFO [train.py:901] (1/4) Epoch 18, batch 3150, loss[loss=0.2457, simple_loss=0.3269, pruned_loss=0.08225, over 8363.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2984, pruned_loss=0.06958, over 1617750.70 frames. ], batch size: 24, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:23:26,371 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140562.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:27,686 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140564.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:34,940 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.438e+02 2.948e+02 4.263e+02 1.019e+03, threshold=5.895e+02, percent-clipped=4.0 +2023-02-06 21:23:38,565 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140580.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:40,650 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:43,430 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:59,140 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:24:01,007 INFO [train.py:901] (1/4) Epoch 18, batch 3200, loss[loss=0.2385, simple_loss=0.3087, pruned_loss=0.0842, over 8460.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2969, pruned_loss=0.06874, over 1617599.34 frames. ], batch size: 27, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:24:06,670 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2404, 2.7277, 3.0636, 1.5754, 3.3102, 2.1223, 1.4844, 2.3515], + device='cuda:1'), covar=tensor([0.0713, 0.0341, 0.0219, 0.0737, 0.0377, 0.0682, 0.0831, 0.0432], + device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0375, 0.0321, 0.0430, 0.0360, 0.0520, 0.0379, 0.0399], + device='cuda:1'), out_proj_covar=tensor([1.1818e-04, 9.9388e-05, 8.5057e-05, 1.1472e-04, 9.6150e-05, 1.4908e-04, + 1.0321e-04, 1.0654e-04], device='cuda:1') +2023-02-06 21:24:07,873 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140621.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:24:09,395 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4787, 2.6909, 1.9615, 2.3245, 2.4007, 1.6992, 2.2458, 2.2396], + device='cuda:1'), covar=tensor([0.1545, 0.0398, 0.1153, 0.0602, 0.0632, 0.1474, 0.0947, 0.1032], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0233, 0.0323, 0.0302, 0.0296, 0.0330, 0.0339, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 21:24:30,744 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3905, 1.6955, 1.7062, 1.0193, 1.7039, 1.3340, 0.2663, 1.6036], + device='cuda:1'), covar=tensor([0.0436, 0.0346, 0.0260, 0.0423, 0.0421, 0.0817, 0.0772, 0.0260], + device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0377, 0.0322, 0.0432, 0.0361, 0.0522, 0.0380, 0.0400], + device='cuda:1'), out_proj_covar=tensor([1.1879e-04, 9.9887e-05, 8.5326e-05, 1.1498e-04, 9.6401e-05, 1.4973e-04, + 1.0349e-04, 1.0701e-04], device='cuda:1') +2023-02-06 21:24:36,814 INFO [train.py:901] (1/4) Epoch 18, batch 3250, loss[loss=0.2099, simple_loss=0.2801, pruned_loss=0.06982, over 7651.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2956, pruned_loss=0.06798, over 1614317.87 frames. ], batch size: 19, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:24:46,453 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.256e+02 2.889e+02 3.448e+02 6.536e+02, threshold=5.777e+02, percent-clipped=1.0 +2023-02-06 21:25:13,081 INFO [train.py:901] (1/4) Epoch 18, batch 3300, loss[loss=0.2453, simple_loss=0.3208, pruned_loss=0.08485, over 8034.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2952, pruned_loss=0.06793, over 1614317.86 frames. ], batch size: 22, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:25:22,137 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1894, 1.3687, 1.7190, 1.3400, 0.6576, 1.4536, 1.2174, 1.0984], + device='cuda:1'), covar=tensor([0.0559, 0.1189, 0.1595, 0.1344, 0.0557, 0.1397, 0.0667, 0.0647], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0100, 0.0161, 0.0113, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 21:25:30,570 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:25:33,922 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140741.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:25:39,313 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140749.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:25:47,455 INFO [train.py:901] (1/4) Epoch 18, batch 3350, loss[loss=0.244, simple_loss=0.3144, pruned_loss=0.08675, over 8697.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2958, pruned_loss=0.06842, over 1615035.10 frames. ], batch size: 34, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:25:57,593 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.504e+02 2.969e+02 3.727e+02 7.020e+02, threshold=5.938e+02, percent-clipped=2.0 +2023-02-06 21:26:02,376 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 21:26:23,938 INFO [train.py:901] (1/4) Epoch 18, batch 3400, loss[loss=0.2075, simple_loss=0.2811, pruned_loss=0.06696, over 8086.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2954, pruned_loss=0.06803, over 1615548.73 frames. ], batch size: 21, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:26:35,805 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140827.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:26:42,133 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140836.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:26:46,749 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140843.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:26:52,076 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:26:58,805 INFO [train.py:901] (1/4) Epoch 18, batch 3450, loss[loss=0.2108, simple_loss=0.2804, pruned_loss=0.0706, over 7706.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2953, pruned_loss=0.06791, over 1611206.96 frames. ], batch size: 18, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:26:59,027 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140861.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:27:01,046 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140864.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:27:03,745 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:27:05,719 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140871.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:27:08,266 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.419e+02 3.065e+02 3.703e+02 6.567e+02, threshold=6.131e+02, percent-clipped=3.0 +2023-02-06 21:27:34,146 INFO [train.py:901] (1/4) Epoch 18, batch 3500, loss[loss=0.1851, simple_loss=0.2489, pruned_loss=0.06069, over 7818.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2957, pruned_loss=0.06804, over 1611794.86 frames. ], batch size: 19, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:27:51,077 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 21:27:51,205 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140935.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:09,212 INFO [train.py:901] (1/4) Epoch 18, batch 3550, loss[loss=0.2252, simple_loss=0.2933, pruned_loss=0.07852, over 7925.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.296, pruned_loss=0.06852, over 1611770.57 frames. ], batch size: 20, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:28:12,011 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:12,766 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140966.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:18,753 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.456e+02 3.083e+02 3.681e+02 6.081e+02, threshold=6.167e+02, percent-clipped=0.0 +2023-02-06 21:28:26,454 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140986.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:30,648 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140992.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:44,277 INFO [train.py:901] (1/4) Epoch 18, batch 3600, loss[loss=0.2158, simple_loss=0.2946, pruned_loss=0.0685, over 8076.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2951, pruned_loss=0.06822, over 1610121.70 frames. ], batch size: 21, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:28:49,258 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141017.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:29:20,372 INFO [train.py:901] (1/4) Epoch 18, batch 3650, loss[loss=0.2454, simple_loss=0.3073, pruned_loss=0.09173, over 7794.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2943, pruned_loss=0.06783, over 1609832.84 frames. ], batch size: 19, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:29:30,812 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.345e+02 2.956e+02 3.633e+02 6.454e+02, threshold=5.912e+02, percent-clipped=1.0 +2023-02-06 21:29:37,716 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141085.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:29:55,728 INFO [train.py:901] (1/4) Epoch 18, batch 3700, loss[loss=0.2894, simple_loss=0.3613, pruned_loss=0.1087, over 7484.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.293, pruned_loss=0.06738, over 1608814.36 frames. ], batch size: 72, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:29:57,129 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 21:30:02,962 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:30:20,682 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141145.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:30:32,113 INFO [train.py:901] (1/4) Epoch 18, batch 3750, loss[loss=0.1871, simple_loss=0.2772, pruned_loss=0.04845, over 8497.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2937, pruned_loss=0.06737, over 1609613.28 frames. ], batch size: 29, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:30:32,261 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:30:39,107 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141171.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:30:41,863 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.679e+02 3.309e+02 4.099e+02 7.455e+02, threshold=6.618e+02, percent-clipped=7.0 +2023-02-06 21:31:00,278 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141200.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:31:07,419 INFO [train.py:901] (1/4) Epoch 18, batch 3800, loss[loss=0.2205, simple_loss=0.2866, pruned_loss=0.07719, over 8086.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2936, pruned_loss=0.06729, over 1610525.20 frames. ], batch size: 21, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:31:15,006 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141222.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:31:29,273 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141242.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:31:32,639 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:31:42,609 INFO [train.py:901] (1/4) Epoch 18, batch 3850, loss[loss=0.2336, simple_loss=0.323, pruned_loss=0.07215, over 8197.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2938, pruned_loss=0.06711, over 1610489.53 frames. ], batch size: 23, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:31:46,905 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141267.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:31:52,713 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.500e+02 3.018e+02 3.684e+02 7.912e+02, threshold=6.036e+02, percent-clipped=1.0 +2023-02-06 21:31:55,474 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141279.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:32:00,514 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:32:03,897 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 21:32:17,056 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141309.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:32:18,347 INFO [train.py:901] (1/4) Epoch 18, batch 3900, loss[loss=0.1882, simple_loss=0.2655, pruned_loss=0.05546, over 7247.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2942, pruned_loss=0.06725, over 1611517.99 frames. ], batch size: 16, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:32:42,386 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:32:51,252 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1202, 1.3126, 3.2207, 1.0382, 2.8435, 2.6784, 2.9370, 2.8874], + device='cuda:1'), covar=tensor([0.0810, 0.3881, 0.0855, 0.4072, 0.1370, 0.1155, 0.0764, 0.0840], + device='cuda:1'), in_proj_covar=tensor([0.0589, 0.0620, 0.0666, 0.0597, 0.0681, 0.0581, 0.0573, 0.0640], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:32:52,480 INFO [train.py:901] (1/4) Epoch 18, batch 3950, loss[loss=0.2231, simple_loss=0.3041, pruned_loss=0.07104, over 8075.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2937, pruned_loss=0.06719, over 1606190.46 frames. ], batch size: 21, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:33:02,714 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.421e+02 2.990e+02 3.795e+02 7.053e+02, threshold=5.979e+02, percent-clipped=3.0 +2023-02-06 21:33:08,820 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9702, 2.5319, 3.5829, 1.6141, 1.9347, 3.6481, 0.6217, 2.1612], + device='cuda:1'), covar=tensor([0.1430, 0.1268, 0.0351, 0.2314, 0.2845, 0.0240, 0.2653, 0.1525], + device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0189, 0.0120, 0.0217, 0.0261, 0.0128, 0.0164, 0.0183], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 21:33:15,883 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141394.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:33:27,610 INFO [train.py:901] (1/4) Epoch 18, batch 4000, loss[loss=0.228, simple_loss=0.3115, pruned_loss=0.07226, over 8453.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2949, pruned_loss=0.0675, over 1610934.88 frames. ], batch size: 27, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:33:34,617 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4184, 1.3628, 2.7690, 1.3781, 2.0193, 2.9835, 3.1409, 2.4647], + device='cuda:1'), covar=tensor([0.1352, 0.1725, 0.0478, 0.2147, 0.1001, 0.0376, 0.0550, 0.0733], + device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0310, 0.0274, 0.0305, 0.0294, 0.0253, 0.0393, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 21:33:37,281 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141424.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:33:44,494 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141435.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:33:53,880 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9182, 1.8412, 3.5207, 1.4809, 2.3729, 3.9395, 3.9846, 3.4336], + device='cuda:1'), covar=tensor([0.1146, 0.1510, 0.0330, 0.1933, 0.1012, 0.0191, 0.0360, 0.0491], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0311, 0.0274, 0.0305, 0.0295, 0.0253, 0.0394, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 21:33:58,735 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141456.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:34:00,782 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141459.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:34:02,057 INFO [train.py:901] (1/4) Epoch 18, batch 4050, loss[loss=0.2592, simple_loss=0.3355, pruned_loss=0.09149, over 8351.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2959, pruned_loss=0.06782, over 1616670.18 frames. ], batch size: 26, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:34:12,709 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.514e+02 3.146e+02 4.229e+02 8.641e+02, threshold=6.293e+02, percent-clipped=9.0 +2023-02-06 21:34:16,954 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141481.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:34:31,499 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8945, 1.7866, 1.9843, 1.8626, 0.9999, 1.7051, 2.3909, 2.4207], + device='cuda:1'), covar=tensor([0.0413, 0.1096, 0.1507, 0.1216, 0.0529, 0.1327, 0.0553, 0.0473], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0157, 0.0099, 0.0162, 0.0114, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 21:34:34,612 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141505.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:34:38,490 INFO [train.py:901] (1/4) Epoch 18, batch 4100, loss[loss=0.2166, simple_loss=0.3002, pruned_loss=0.06645, over 8601.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2956, pruned_loss=0.06736, over 1619586.01 frames. ], batch size: 34, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:35:00,220 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141542.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:35:13,075 INFO [train.py:901] (1/4) Epoch 18, batch 4150, loss[loss=0.1918, simple_loss=0.2701, pruned_loss=0.0568, over 7977.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2953, pruned_loss=0.06661, over 1619626.44 frames. ], batch size: 21, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:35:17,351 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141567.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:35:22,679 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.507e+02 2.964e+02 3.952e+02 7.900e+02, threshold=5.928e+02, percent-clipped=3.0 +2023-02-06 21:35:48,917 INFO [train.py:901] (1/4) Epoch 18, batch 4200, loss[loss=0.1974, simple_loss=0.2742, pruned_loss=0.06023, over 8084.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2943, pruned_loss=0.06624, over 1618830.47 frames. ], batch size: 21, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:35:55,109 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:02,352 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 21:36:15,991 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141650.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:23,890 INFO [train.py:901] (1/4) Epoch 18, batch 4250, loss[loss=0.1932, simple_loss=0.2662, pruned_loss=0.0601, over 7715.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2937, pruned_loss=0.06602, over 1616229.44 frames. ], batch size: 18, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:36:24,588 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 21:36:33,264 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.491e+02 2.994e+02 3.932e+02 8.485e+02, threshold=5.988e+02, percent-clipped=6.0 +2023-02-06 21:36:33,474 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141675.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:36,880 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141680.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:44,000 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:54,152 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141705.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:58,037 INFO [train.py:901] (1/4) Epoch 18, batch 4300, loss[loss=0.2277, simple_loss=0.3085, pruned_loss=0.07348, over 8506.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2942, pruned_loss=0.0661, over 1618267.85 frames. ], batch size: 26, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:37:01,163 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.47 vs. limit=5.0 +2023-02-06 21:37:32,940 INFO [train.py:901] (1/4) Epoch 18, batch 4350, loss[loss=0.2244, simple_loss=0.3027, pruned_loss=0.073, over 8383.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2937, pruned_loss=0.06619, over 1613702.34 frames. ], batch size: 49, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:37:43,209 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 2.620e+02 3.197e+02 4.150e+02 9.266e+02, threshold=6.393e+02, percent-clipped=5.0 +2023-02-06 21:37:46,110 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141779.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:37:54,229 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 21:38:02,428 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141803.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:38:04,519 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141806.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:38:07,892 INFO [train.py:901] (1/4) Epoch 18, batch 4400, loss[loss=0.2047, simple_loss=0.2855, pruned_loss=0.06193, over 8614.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.294, pruned_loss=0.06635, over 1613107.79 frames. ], batch size: 39, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:38:36,597 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 21:38:43,844 INFO [train.py:901] (1/4) Epoch 18, batch 4450, loss[loss=0.1732, simple_loss=0.2417, pruned_loss=0.05241, over 7431.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.294, pruned_loss=0.06723, over 1608084.55 frames. ], batch size: 17, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:38:53,329 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.507e+02 2.868e+02 3.524e+02 7.777e+02, threshold=5.735e+02, percent-clipped=2.0 +2023-02-06 21:38:54,254 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:38:56,857 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2025, 3.1242, 2.8937, 1.5298, 2.8200, 2.8556, 2.8438, 2.7137], + device='cuda:1'), covar=tensor([0.1293, 0.0980, 0.1411, 0.4931, 0.1184, 0.1394, 0.1751, 0.1125], + device='cuda:1'), in_proj_covar=tensor([0.0508, 0.0423, 0.0422, 0.0522, 0.0414, 0.0420, 0.0405, 0.0369], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:39:07,111 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141894.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:39:11,825 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141901.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:39:18,281 INFO [train.py:901] (1/4) Epoch 18, batch 4500, loss[loss=0.2618, simple_loss=0.3474, pruned_loss=0.08804, over 8436.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2939, pruned_loss=0.06669, over 1610791.79 frames. ], batch size: 27, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:39:23,246 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141918.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:39:26,144 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.00 vs. limit=5.0 +2023-02-06 21:39:27,816 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 21:39:34,712 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141934.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:39:42,681 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141946.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:39:53,392 INFO [train.py:901] (1/4) Epoch 18, batch 4550, loss[loss=0.2389, simple_loss=0.3153, pruned_loss=0.08122, over 8452.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2944, pruned_loss=0.0674, over 1608909.85 frames. ], batch size: 27, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:40:03,507 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.488e+02 2.920e+02 3.454e+02 6.371e+02, threshold=5.840e+02, percent-clipped=2.0 +2023-02-06 21:40:29,735 INFO [train.py:901] (1/4) Epoch 18, batch 4600, loss[loss=0.2156, simple_loss=0.2965, pruned_loss=0.06734, over 8469.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2942, pruned_loss=0.06696, over 1612953.31 frames. ], batch size: 29, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:40:31,401 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1868, 1.8971, 2.6372, 2.2470, 2.5868, 2.1965, 1.9223, 1.3123], + device='cuda:1'), covar=tensor([0.4871, 0.4694, 0.1756, 0.3222, 0.2214, 0.2797, 0.1868, 0.4931], + device='cuda:1'), in_proj_covar=tensor([0.0929, 0.0946, 0.0783, 0.0911, 0.0982, 0.0863, 0.0729, 0.0810], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:40:59,842 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.74 vs. limit=5.0 +2023-02-06 21:41:04,190 INFO [train.py:901] (1/4) Epoch 18, batch 4650, loss[loss=0.234, simple_loss=0.3114, pruned_loss=0.0783, over 8523.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2938, pruned_loss=0.0666, over 1618720.46 frames. ], batch size: 28, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:41:05,114 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:41:06,000 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 21:41:13,898 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.389e+02 2.901e+02 3.503e+02 7.256e+02, threshold=5.801e+02, percent-clipped=3.0 +2023-02-06 21:41:23,653 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:41:39,476 INFO [train.py:901] (1/4) Epoch 18, batch 4700, loss[loss=0.1858, simple_loss=0.2819, pruned_loss=0.0449, over 8252.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2948, pruned_loss=0.06652, over 1619325.70 frames. ], batch size: 24, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:42:06,019 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142150.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:42:13,365 INFO [train.py:901] (1/4) Epoch 18, batch 4750, loss[loss=0.216, simple_loss=0.2857, pruned_loss=0.07316, over 7247.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2946, pruned_loss=0.06653, over 1618119.65 frames. ], batch size: 16, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:42:23,433 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142174.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:42:23,892 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.398e+02 2.792e+02 3.541e+02 9.190e+02, threshold=5.585e+02, percent-clipped=4.0 +2023-02-06 21:42:24,096 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142175.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:42:30,586 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 21:42:32,638 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 21:42:41,673 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142199.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:42:49,355 INFO [train.py:901] (1/4) Epoch 18, batch 4800, loss[loss=0.2196, simple_loss=0.308, pruned_loss=0.06556, over 8480.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2946, pruned_loss=0.06645, over 1618061.27 frames. ], batch size: 27, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:43:23,853 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 21:43:24,504 INFO [train.py:901] (1/4) Epoch 18, batch 4850, loss[loss=0.1961, simple_loss=0.2849, pruned_loss=0.05364, over 8678.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2944, pruned_loss=0.06642, over 1616960.35 frames. ], batch size: 34, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:43:33,973 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.520e+02 2.522e+02 3.053e+02 3.876e+02 6.315e+02, threshold=6.106e+02, percent-clipped=2.0 +2023-02-06 21:43:36,094 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=142278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:43:45,102 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=142290.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:43:59,068 INFO [train.py:901] (1/4) Epoch 18, batch 4900, loss[loss=0.2361, simple_loss=0.299, pruned_loss=0.08662, over 7974.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2945, pruned_loss=0.06645, over 1617522.54 frames. ], batch size: 21, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:44:28,577 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1673, 4.1278, 3.8241, 1.9651, 3.7148, 3.7622, 3.7424, 3.5984], + device='cuda:1'), covar=tensor([0.0888, 0.0658, 0.1149, 0.4821, 0.0957, 0.1094, 0.1294, 0.0887], + device='cuda:1'), in_proj_covar=tensor([0.0509, 0.0423, 0.0421, 0.0521, 0.0412, 0.0418, 0.0402, 0.0369], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:44:34,495 INFO [train.py:901] (1/4) Epoch 18, batch 4950, loss[loss=0.2207, simple_loss=0.3143, pruned_loss=0.06356, over 8340.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2947, pruned_loss=0.06668, over 1615213.63 frames. ], batch size: 25, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:44:43,723 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5911, 1.3239, 4.7366, 1.8137, 4.1764, 3.9786, 4.2899, 4.1509], + device='cuda:1'), covar=tensor([0.0528, 0.4772, 0.0480, 0.3890, 0.1083, 0.0972, 0.0512, 0.0631], + device='cuda:1'), in_proj_covar=tensor([0.0588, 0.0620, 0.0665, 0.0597, 0.0676, 0.0579, 0.0571, 0.0639], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:44:44,958 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.451e+02 2.943e+02 3.789e+02 7.945e+02, threshold=5.886e+02, percent-clipped=1.0 +2023-02-06 21:44:57,222 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142393.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:45:05,906 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142405.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:45:07,281 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6865, 1.6759, 2.2277, 1.5746, 1.1908, 2.2583, 0.3509, 1.3604], + device='cuda:1'), covar=tensor([0.1982, 0.1445, 0.0404, 0.1346, 0.3230, 0.0449, 0.2472, 0.1455], + device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0190, 0.0121, 0.0218, 0.0266, 0.0129, 0.0165, 0.0184], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 21:45:09,741 INFO [train.py:901] (1/4) Epoch 18, batch 5000, loss[loss=0.2378, simple_loss=0.3051, pruned_loss=0.08521, over 7516.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2947, pruned_loss=0.06686, over 1614229.91 frames. ], batch size: 18, lr: 4.21e-03, grad_scale: 8.0 +2023-02-06 21:45:44,307 INFO [train.py:901] (1/4) Epoch 18, batch 5050, loss[loss=0.2217, simple_loss=0.2996, pruned_loss=0.07192, over 8459.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2954, pruned_loss=0.06685, over 1622744.65 frames. ], batch size: 29, lr: 4.21e-03, grad_scale: 8.0 +2023-02-06 21:45:54,481 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.384e+02 2.804e+02 3.417e+02 5.925e+02, threshold=5.609e+02, percent-clipped=1.0 +2023-02-06 21:46:04,051 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 21:46:06,029 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-06 21:46:19,171 INFO [train.py:901] (1/4) Epoch 18, batch 5100, loss[loss=0.2177, simple_loss=0.2911, pruned_loss=0.07217, over 8497.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2959, pruned_loss=0.06771, over 1615039.34 frames. ], batch size: 26, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:46:54,248 INFO [train.py:901] (1/4) Epoch 18, batch 5150, loss[loss=0.2382, simple_loss=0.3179, pruned_loss=0.07923, over 8474.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2954, pruned_loss=0.06728, over 1618058.22 frames. ], batch size: 25, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:47:04,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.481e+02 3.004e+02 4.323e+02 1.197e+03, threshold=6.009e+02, percent-clipped=7.0 +2023-02-06 21:47:07,363 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7001, 2.7427, 2.0161, 2.4940, 2.4084, 1.7952, 2.2809, 2.3468], + device='cuda:1'), covar=tensor([0.1388, 0.0422, 0.1105, 0.0588, 0.0690, 0.1360, 0.0927, 0.0977], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0237, 0.0327, 0.0306, 0.0300, 0.0332, 0.0346, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 21:47:14,692 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142591.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:47:29,005 INFO [train.py:901] (1/4) Epoch 18, batch 5200, loss[loss=0.192, simple_loss=0.2711, pruned_loss=0.05641, over 7667.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2952, pruned_loss=0.06721, over 1615244.92 frames. ], batch size: 19, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:47:37,468 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2639, 1.8622, 2.4046, 2.0066, 2.2459, 2.1995, 1.9606, 1.1297], + device='cuda:1'), covar=tensor([0.4631, 0.4203, 0.1712, 0.3148, 0.2321, 0.2681, 0.1783, 0.4788], + device='cuda:1'), in_proj_covar=tensor([0.0926, 0.0942, 0.0778, 0.0908, 0.0980, 0.0861, 0.0731, 0.0809], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:47:55,319 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:48:02,984 INFO [train.py:901] (1/4) Epoch 18, batch 5250, loss[loss=0.223, simple_loss=0.3062, pruned_loss=0.06991, over 8327.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2949, pruned_loss=0.06717, over 1615862.92 frames. ], batch size: 25, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:48:03,213 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:48:03,671 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 21:48:11,774 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1111, 1.3571, 4.3122, 1.6494, 3.7596, 3.5577, 3.8237, 3.7490], + device='cuda:1'), covar=tensor([0.0660, 0.4763, 0.0509, 0.3948, 0.1185, 0.0967, 0.0712, 0.0670], + device='cuda:1'), in_proj_covar=tensor([0.0589, 0.0618, 0.0666, 0.0594, 0.0674, 0.0579, 0.0572, 0.0639], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:48:13,216 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142674.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:48:14,366 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.428e+02 2.491e+02 3.102e+02 3.692e+02 6.533e+02, threshold=6.204e+02, percent-clipped=2.0 +2023-02-06 21:48:21,233 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142686.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:48:37,892 INFO [train.py:901] (1/4) Epoch 18, batch 5300, loss[loss=0.2409, simple_loss=0.329, pruned_loss=0.07641, over 8529.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2951, pruned_loss=0.06735, over 1619472.89 frames. ], batch size: 28, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:48:39,192 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 21:48:52,079 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4705, 1.8990, 2.9290, 1.3197, 2.1770, 1.8502, 1.6085, 2.0700], + device='cuda:1'), covar=tensor([0.2141, 0.2666, 0.0966, 0.5000, 0.1972, 0.3547, 0.2499, 0.2546], + device='cuda:1'), in_proj_covar=tensor([0.0514, 0.0577, 0.0550, 0.0620, 0.0635, 0.0584, 0.0513, 0.0627], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:48:59,489 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142742.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:49:12,896 INFO [train.py:901] (1/4) Epoch 18, batch 5350, loss[loss=0.2501, simple_loss=0.3188, pruned_loss=0.09068, over 7287.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2949, pruned_loss=0.06735, over 1618853.87 frames. ], batch size: 73, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:49:22,771 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.581e+02 3.011e+02 3.651e+02 7.168e+02, threshold=6.023e+02, percent-clipped=3.0 +2023-02-06 21:49:48,114 INFO [train.py:901] (1/4) Epoch 18, batch 5400, loss[loss=0.1871, simple_loss=0.2639, pruned_loss=0.05515, over 7440.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2935, pruned_loss=0.06667, over 1611998.49 frames. ], batch size: 17, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:50:11,410 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5568, 4.5244, 4.1385, 1.9105, 4.0152, 4.1555, 4.0813, 4.0146], + device='cuda:1'), covar=tensor([0.0665, 0.0501, 0.0866, 0.4716, 0.0855, 0.0866, 0.1153, 0.0730], + device='cuda:1'), in_proj_covar=tensor([0.0507, 0.0422, 0.0420, 0.0521, 0.0412, 0.0420, 0.0404, 0.0369], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:50:22,756 INFO [train.py:901] (1/4) Epoch 18, batch 5450, loss[loss=0.216, simple_loss=0.3074, pruned_loss=0.0623, over 8357.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2939, pruned_loss=0.06634, over 1617273.12 frames. ], batch size: 24, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:50:33,556 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 2.381e+02 3.003e+02 4.378e+02 7.690e+02, threshold=6.006e+02, percent-clipped=4.0 +2023-02-06 21:50:33,771 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2747, 2.6508, 2.2030, 3.6176, 1.8034, 2.0879, 2.1976, 2.8763], + device='cuda:1'), covar=tensor([0.0713, 0.0761, 0.0881, 0.0339, 0.1052, 0.1138, 0.0995, 0.0708], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0199, 0.0252, 0.0213, 0.0207, 0.0247, 0.0254, 0.0212], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 21:50:44,189 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 +2023-02-06 21:50:44,889 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-02-06 21:50:50,020 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 21:50:51,828 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 21:50:58,892 INFO [train.py:901] (1/4) Epoch 18, batch 5500, loss[loss=0.209, simple_loss=0.2919, pruned_loss=0.06303, over 8245.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2934, pruned_loss=0.06584, over 1617451.82 frames. ], batch size: 22, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:51:00,980 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6910, 2.0926, 3.2751, 1.5099, 2.5211, 2.0673, 1.8597, 2.4815], + device='cuda:1'), covar=tensor([0.1753, 0.2323, 0.0800, 0.4127, 0.1800, 0.2944, 0.1998, 0.2221], + device='cuda:1'), in_proj_covar=tensor([0.0510, 0.0572, 0.0548, 0.0616, 0.0631, 0.0579, 0.0510, 0.0623], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:51:08,975 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9321, 1.4699, 1.7857, 1.5106, 1.0423, 1.5292, 2.2933, 2.0009], + device='cuda:1'), covar=tensor([0.0450, 0.1354, 0.1804, 0.1494, 0.0619, 0.1589, 0.0626, 0.0622], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0191, 0.0158, 0.0100, 0.0162, 0.0114, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 21:51:14,989 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=142935.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:51:33,201 INFO [train.py:901] (1/4) Epoch 18, batch 5550, loss[loss=0.1915, simple_loss=0.2807, pruned_loss=0.0512, over 8185.00 frames. ], tot_loss[loss=0.212, simple_loss=0.293, pruned_loss=0.06544, over 1620006.28 frames. ], batch size: 23, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:51:43,334 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.398e+02 2.938e+02 3.826e+02 1.126e+03, threshold=5.876e+02, percent-clipped=10.0 +2023-02-06 21:52:08,271 INFO [train.py:901] (1/4) Epoch 18, batch 5600, loss[loss=0.2455, simple_loss=0.325, pruned_loss=0.08307, over 8576.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2932, pruned_loss=0.06585, over 1614363.35 frames. ], batch size: 31, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:52:36,265 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143050.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:52:43,528 INFO [train.py:901] (1/4) Epoch 18, batch 5650, loss[loss=0.1802, simple_loss=0.2547, pruned_loss=0.05283, over 7701.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2929, pruned_loss=0.06558, over 1615071.36 frames. ], batch size: 18, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:52:54,554 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.315e+02 3.071e+02 3.627e+02 7.364e+02, threshold=6.141e+02, percent-clipped=4.0 +2023-02-06 21:53:00,418 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 21:53:01,153 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=143086.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:53:18,751 INFO [train.py:901] (1/4) Epoch 18, batch 5700, loss[loss=0.1982, simple_loss=0.2907, pruned_loss=0.05283, over 8721.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2935, pruned_loss=0.06614, over 1616186.70 frames. ], batch size: 39, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:53:21,077 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.92 vs. limit=5.0 +2023-02-06 21:53:53,727 INFO [train.py:901] (1/4) Epoch 18, batch 5750, loss[loss=0.2454, simple_loss=0.335, pruned_loss=0.07796, over 8487.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2927, pruned_loss=0.06594, over 1614514.83 frames. ], batch size: 27, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:54:04,019 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.527e+02 2.423e+02 2.839e+02 3.621e+02 5.889e+02, threshold=5.677e+02, percent-clipped=0.0 +2023-02-06 21:54:04,725 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 21:54:21,874 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143201.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:54:28,669 INFO [train.py:901] (1/4) Epoch 18, batch 5800, loss[loss=0.2235, simple_loss=0.2958, pruned_loss=0.07564, over 7807.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2932, pruned_loss=0.06628, over 1610795.81 frames. ], batch size: 20, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:54:29,535 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5413, 1.3030, 4.7697, 1.7765, 4.2469, 3.9877, 4.2778, 4.1799], + device='cuda:1'), covar=tensor([0.0587, 0.4427, 0.0450, 0.3792, 0.0993, 0.0977, 0.0549, 0.0575], + device='cuda:1'), in_proj_covar=tensor([0.0595, 0.0623, 0.0672, 0.0604, 0.0680, 0.0584, 0.0576, 0.0646], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 21:54:59,278 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0354, 2.3363, 3.4970, 1.8786, 2.9612, 2.4068, 2.1905, 2.7976], + device='cuda:1'), covar=tensor([0.1515, 0.2215, 0.0800, 0.3660, 0.1460, 0.2494, 0.1826, 0.2125], + device='cuda:1'), in_proj_covar=tensor([0.0510, 0.0572, 0.0547, 0.0618, 0.0631, 0.0577, 0.0511, 0.0622], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 21:55:04,397 INFO [train.py:901] (1/4) Epoch 18, batch 5850, loss[loss=0.177, simple_loss=0.2542, pruned_loss=0.04988, over 7419.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2939, pruned_loss=0.06635, over 1614830.11 frames. ], batch size: 17, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:55:15,560 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 2.467e+02 2.892e+02 3.630e+02 6.628e+02, threshold=5.783e+02, percent-clipped=2.0 +2023-02-06 21:55:21,404 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.65 vs. limit=5.0 +2023-02-06 21:55:25,402 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 21:55:36,559 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143306.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:55:39,803 INFO [train.py:901] (1/4) Epoch 18, batch 5900, loss[loss=0.2015, simple_loss=0.2793, pruned_loss=0.06185, over 8085.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2933, pruned_loss=0.06603, over 1611908.69 frames. ], batch size: 21, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:55:53,322 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143331.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:56:10,806 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 21:56:14,464 INFO [train.py:901] (1/4) Epoch 18, batch 5950, loss[loss=0.2257, simple_loss=0.3055, pruned_loss=0.07297, over 8373.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.294, pruned_loss=0.06665, over 1612195.92 frames. ], batch size: 24, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:56:25,158 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.388e+02 2.875e+02 3.741e+02 7.794e+02, threshold=5.749e+02, percent-clipped=3.0 +2023-02-06 21:56:37,556 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3803, 2.5417, 1.8011, 2.2361, 2.0713, 1.5086, 1.9692, 2.0882], + device='cuda:1'), covar=tensor([0.1467, 0.0449, 0.1189, 0.0622, 0.0788, 0.1540, 0.1005, 0.0850], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0236, 0.0328, 0.0306, 0.0298, 0.0333, 0.0344, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 21:56:49,478 INFO [train.py:901] (1/4) Epoch 18, batch 6000, loss[loss=0.2076, simple_loss=0.2867, pruned_loss=0.06421, over 7234.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2936, pruned_loss=0.06635, over 1613417.97 frames. ], batch size: 16, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:56:49,478 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 21:57:03,436 INFO [train.py:935] (1/4) Epoch 18, validation: loss=0.1765, simple_loss=0.2767, pruned_loss=0.03814, over 944034.00 frames. +2023-02-06 21:57:03,437 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 21:57:08,526 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143418.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:57:35,797 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143457.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:57:38,431 INFO [train.py:901] (1/4) Epoch 18, batch 6050, loss[loss=0.2134, simple_loss=0.3002, pruned_loss=0.06332, over 8489.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2935, pruned_loss=0.06619, over 1610894.80 frames. ], batch size: 26, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:57:48,577 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.412e+02 3.060e+02 4.409e+02 1.030e+03, threshold=6.120e+02, percent-clipped=9.0 +2023-02-06 21:57:53,572 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143482.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:57:59,600 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143491.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 21:58:13,406 INFO [train.py:901] (1/4) Epoch 18, batch 6100, loss[loss=0.191, simple_loss=0.2718, pruned_loss=0.05513, over 8041.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.294, pruned_loss=0.06607, over 1618671.38 frames. ], batch size: 22, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:58:39,427 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 21:58:49,839 INFO [train.py:901] (1/4) Epoch 18, batch 6150, loss[loss=0.1919, simple_loss=0.2712, pruned_loss=0.05626, over 8023.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2936, pruned_loss=0.06612, over 1616678.68 frames. ], batch size: 22, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:59:00,212 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.359e+02 3.030e+02 3.820e+02 7.737e+02, threshold=6.061e+02, percent-clipped=3.0 +2023-02-06 21:59:02,577 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0755, 1.0148, 1.2935, 0.9806, 0.8484, 1.2945, 0.2955, 0.9567], + device='cuda:1'), covar=tensor([0.1616, 0.0992, 0.0360, 0.0782, 0.2282, 0.0467, 0.1916, 0.1169], + device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0188, 0.0120, 0.0216, 0.0263, 0.0128, 0.0165, 0.0182], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 21:59:25,641 INFO [train.py:901] (1/4) Epoch 18, batch 6200, loss[loss=0.2152, simple_loss=0.2942, pruned_loss=0.06806, over 8240.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2935, pruned_loss=0.06601, over 1618412.55 frames. ], batch size: 22, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:59:33,676 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0894, 3.0793, 2.2254, 2.5109, 2.4772, 2.1128, 2.3860, 2.6696], + device='cuda:1'), covar=tensor([0.1224, 0.0310, 0.0848, 0.0634, 0.0576, 0.1098, 0.0799, 0.0823], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0234, 0.0325, 0.0305, 0.0296, 0.0330, 0.0342, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 22:00:01,324 INFO [train.py:901] (1/4) Epoch 18, batch 6250, loss[loss=0.1823, simple_loss=0.2659, pruned_loss=0.04931, over 7967.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2938, pruned_loss=0.06607, over 1620172.73 frames. ], batch size: 21, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 22:00:12,414 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.460e+02 3.089e+02 4.040e+02 1.017e+03, threshold=6.178e+02, percent-clipped=5.0 +2023-02-06 22:00:37,038 INFO [train.py:901] (1/4) Epoch 18, batch 6300, loss[loss=0.3063, simple_loss=0.3568, pruned_loss=0.1279, over 7197.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2939, pruned_loss=0.06608, over 1620259.87 frames. ], batch size: 72, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 22:00:38,579 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5558, 4.5609, 4.1143, 1.9879, 4.0292, 4.1901, 4.1468, 3.9054], + device='cuda:1'), covar=tensor([0.0725, 0.0575, 0.1114, 0.4948, 0.0886, 0.0920, 0.1382, 0.0773], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0425, 0.0425, 0.0524, 0.0415, 0.0421, 0.0408, 0.0374], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:00:54,435 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1830, 2.5396, 2.8659, 1.5676, 3.1499, 1.8065, 1.5076, 2.0414], + device='cuda:1'), covar=tensor([0.0720, 0.0324, 0.0254, 0.0694, 0.0364, 0.0802, 0.0895, 0.0470], + device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0372, 0.0321, 0.0432, 0.0361, 0.0522, 0.0378, 0.0397], + device='cuda:1'), out_proj_covar=tensor([1.1797e-04, 9.8280e-05, 8.5080e-05, 1.1498e-04, 9.6080e-05, 1.4946e-04, + 1.0271e-04, 1.0566e-04], device='cuda:1') +2023-02-06 22:01:11,904 INFO [train.py:901] (1/4) Epoch 18, batch 6350, loss[loss=0.1885, simple_loss=0.2728, pruned_loss=0.0521, over 7939.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2934, pruned_loss=0.06569, over 1619247.58 frames. ], batch size: 20, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 22:01:13,343 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=143762.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:01:22,540 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.204e+02 2.882e+02 3.589e+02 6.333e+02, threshold=5.763e+02, percent-clipped=1.0 +2023-02-06 22:01:47,364 INFO [train.py:901] (1/4) Epoch 18, batch 6400, loss[loss=0.2311, simple_loss=0.3036, pruned_loss=0.07925, over 7786.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2939, pruned_loss=0.06648, over 1614098.41 frames. ], batch size: 19, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 22:02:04,437 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=143835.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:02:21,085 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-06 22:02:21,890 INFO [train.py:901] (1/4) Epoch 18, batch 6450, loss[loss=0.1952, simple_loss=0.2771, pruned_loss=0.05666, over 7817.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2939, pruned_loss=0.06629, over 1616056.83 frames. ], batch size: 20, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:02:33,448 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.452e+02 2.973e+02 3.704e+02 1.405e+03, threshold=5.946e+02, percent-clipped=1.0 +2023-02-06 22:02:34,290 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143877.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:02:57,286 INFO [train.py:901] (1/4) Epoch 18, batch 6500, loss[loss=0.22, simple_loss=0.3125, pruned_loss=0.06374, over 8501.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.295, pruned_loss=0.06729, over 1618828.19 frames. ], batch size: 26, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:03:24,076 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143950.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:03:31,368 INFO [train.py:901] (1/4) Epoch 18, batch 6550, loss[loss=0.2331, simple_loss=0.3171, pruned_loss=0.07453, over 8477.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2952, pruned_loss=0.06764, over 1620034.01 frames. ], batch size: 29, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:03:41,858 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.926e+02 2.526e+02 3.154e+02 3.765e+02 8.734e+02, threshold=6.308e+02, percent-clipped=5.0 +2023-02-06 22:03:48,839 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 22:04:08,871 INFO [train.py:901] (1/4) Epoch 18, batch 6600, loss[loss=0.1714, simple_loss=0.2546, pruned_loss=0.04416, over 7542.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2953, pruned_loss=0.06763, over 1619461.80 frames. ], batch size: 18, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:04:10,899 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 22:04:37,996 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 22:04:43,667 INFO [train.py:901] (1/4) Epoch 18, batch 6650, loss[loss=0.211, simple_loss=0.2973, pruned_loss=0.06234, over 8340.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2947, pruned_loss=0.06718, over 1619671.99 frames. ], batch size: 26, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:04:54,721 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.649e+02 2.298e+02 3.022e+02 3.555e+02 7.360e+02, threshold=6.043e+02, percent-clipped=4.0 +2023-02-06 22:05:19,671 INFO [train.py:901] (1/4) Epoch 18, batch 6700, loss[loss=0.176, simple_loss=0.2642, pruned_loss=0.04391, over 7972.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2937, pruned_loss=0.06651, over 1617167.42 frames. ], batch size: 21, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:05:34,397 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144133.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:05:51,957 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144158.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:05:53,839 INFO [train.py:901] (1/4) Epoch 18, batch 6750, loss[loss=0.1799, simple_loss=0.2682, pruned_loss=0.04581, over 8621.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2934, pruned_loss=0.06645, over 1615664.02 frames. ], batch size: 39, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:06:00,259 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 22:06:03,938 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.293e+02 3.003e+02 3.717e+02 7.578e+02, threshold=6.007e+02, percent-clipped=1.0 +2023-02-06 22:06:09,033 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9983, 2.2512, 1.8983, 2.7706, 1.3255, 1.6627, 1.9700, 2.2444], + device='cuda:1'), covar=tensor([0.0776, 0.0727, 0.0998, 0.0416, 0.1127, 0.1341, 0.0933, 0.0761], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0200, 0.0253, 0.0214, 0.0209, 0.0250, 0.0254, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 22:06:17,309 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6185, 1.8345, 2.0360, 1.3158, 2.0999, 1.4727, 0.5127, 1.8077], + device='cuda:1'), covar=tensor([0.0538, 0.0354, 0.0236, 0.0525, 0.0358, 0.0868, 0.0799, 0.0301], + device='cuda:1'), in_proj_covar=tensor([0.0440, 0.0377, 0.0325, 0.0437, 0.0366, 0.0526, 0.0382, 0.0403], + device='cuda:1'), out_proj_covar=tensor([1.1951e-04, 9.9599e-05, 8.6145e-05, 1.1622e-04, 9.7435e-05, 1.5044e-04, + 1.0377e-04, 1.0759e-04], device='cuda:1') +2023-02-06 22:06:25,446 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144206.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:06:27,267 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 22:06:28,568 INFO [train.py:901] (1/4) Epoch 18, batch 6800, loss[loss=0.2034, simple_loss=0.2874, pruned_loss=0.05972, over 8611.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2919, pruned_loss=0.06583, over 1614929.57 frames. ], batch size: 49, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:06:31,424 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3196, 1.9187, 1.3456, 3.0212, 1.4059, 1.1681, 2.0777, 2.0112], + device='cuda:1'), covar=tensor([0.1702, 0.1141, 0.2191, 0.0379, 0.1270, 0.2069, 0.0929, 0.0927], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0200, 0.0252, 0.0213, 0.0208, 0.0249, 0.0253, 0.0212], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 22:06:42,854 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144231.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:07:03,971 INFO [train.py:901] (1/4) Epoch 18, batch 6850, loss[loss=0.2319, simple_loss=0.3098, pruned_loss=0.07705, over 8497.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2932, pruned_loss=0.06645, over 1613581.22 frames. ], batch size: 26, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:07:13,993 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.379e+02 2.937e+02 3.634e+02 6.722e+02, threshold=5.873e+02, percent-clipped=2.0 +2023-02-06 22:07:17,336 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 22:07:25,243 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-06 22:07:38,061 INFO [train.py:901] (1/4) Epoch 18, batch 6900, loss[loss=0.2327, simple_loss=0.297, pruned_loss=0.08416, over 7716.00 frames. ], tot_loss[loss=0.214, simple_loss=0.294, pruned_loss=0.06694, over 1616267.56 frames. ], batch size: 18, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:08:13,482 INFO [train.py:901] (1/4) Epoch 18, batch 6950, loss[loss=0.2064, simple_loss=0.294, pruned_loss=0.05933, over 8045.00 frames. ], tot_loss[loss=0.214, simple_loss=0.294, pruned_loss=0.06699, over 1616876.73 frames. ], batch size: 22, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:08:21,217 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 22:08:24,083 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.398e+02 2.919e+02 3.864e+02 7.610e+02, threshold=5.839e+02, percent-clipped=3.0 +2023-02-06 22:08:25,486 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 22:08:47,777 INFO [train.py:901] (1/4) Epoch 18, batch 7000, loss[loss=0.1583, simple_loss=0.2467, pruned_loss=0.03491, over 7799.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2925, pruned_loss=0.06674, over 1607183.52 frames. ], batch size: 19, lr: 4.18e-03, grad_scale: 16.0 +2023-02-06 22:09:01,099 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144429.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:09:22,580 INFO [train.py:901] (1/4) Epoch 18, batch 7050, loss[loss=0.2136, simple_loss=0.2918, pruned_loss=0.06771, over 7437.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2925, pruned_loss=0.06668, over 1607673.40 frames. ], batch size: 17, lr: 4.18e-03, grad_scale: 16.0 +2023-02-06 22:09:32,284 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1334, 1.2872, 1.6316, 1.2485, 0.7123, 1.3929, 1.2278, 1.0185], + device='cuda:1'), covar=tensor([0.0638, 0.1278, 0.1605, 0.1501, 0.0568, 0.1437, 0.0695, 0.0733], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0099, 0.0162, 0.0113, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 22:09:34,230 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.214e+02 2.534e+02 2.937e+02 3.689e+02 8.247e+02, threshold=5.874e+02, percent-clipped=3.0 +2023-02-06 22:09:37,836 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6052, 2.0947, 3.2625, 1.4300, 2.5439, 2.0867, 1.7640, 2.4252], + device='cuda:1'), covar=tensor([0.1804, 0.2440, 0.0857, 0.4357, 0.1884, 0.3116, 0.2092, 0.2398], + device='cuda:1'), in_proj_covar=tensor([0.0512, 0.0577, 0.0549, 0.0620, 0.0637, 0.0581, 0.0513, 0.0626], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:09:58,439 INFO [train.py:901] (1/4) Epoch 18, batch 7100, loss[loss=0.2204, simple_loss=0.2979, pruned_loss=0.07143, over 8339.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2938, pruned_loss=0.06696, over 1609205.59 frames. ], batch size: 26, lr: 4.18e-03, grad_scale: 16.0 +2023-02-06 22:10:12,489 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9787, 6.0214, 5.3409, 2.8596, 5.3553, 5.7367, 5.6151, 5.4055], + device='cuda:1'), covar=tensor([0.0447, 0.0330, 0.0859, 0.3632, 0.0710, 0.0650, 0.0989, 0.0482], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0426, 0.0428, 0.0528, 0.0418, 0.0423, 0.0411, 0.0372], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:10:30,417 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9631, 3.7430, 2.3495, 2.8885, 2.7341, 2.0802, 2.7706, 2.9235], + device='cuda:1'), covar=tensor([0.1721, 0.0379, 0.0991, 0.0778, 0.0696, 0.1314, 0.1004, 0.0979], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0233, 0.0324, 0.0303, 0.0296, 0.0330, 0.0341, 0.0314], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 22:10:33,593 INFO [train.py:901] (1/4) Epoch 18, batch 7150, loss[loss=0.2328, simple_loss=0.3111, pruned_loss=0.07725, over 8453.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2939, pruned_loss=0.0672, over 1606946.04 frames. ], batch size: 27, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:10:43,979 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.496e+02 2.263e+02 2.906e+02 3.662e+02 1.305e+03, threshold=5.813e+02, percent-clipped=7.0 +2023-02-06 22:11:10,033 INFO [train.py:901] (1/4) Epoch 18, batch 7200, loss[loss=0.2336, simple_loss=0.2995, pruned_loss=0.08383, over 8303.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2934, pruned_loss=0.06681, over 1610075.55 frames. ], batch size: 23, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:11:24,807 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1890, 1.3653, 1.6505, 1.2680, 0.7657, 1.4026, 1.2224, 1.0709], + device='cuda:1'), covar=tensor([0.0568, 0.1225, 0.1637, 0.1431, 0.0550, 0.1423, 0.0677, 0.0678], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0151, 0.0190, 0.0157, 0.0099, 0.0162, 0.0113, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 22:11:35,702 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0300, 2.2145, 1.8615, 2.7114, 1.3721, 1.6902, 1.8409, 2.3145], + device='cuda:1'), covar=tensor([0.0633, 0.0724, 0.0831, 0.0331, 0.1093, 0.1158, 0.0908, 0.0666], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0197, 0.0247, 0.0209, 0.0205, 0.0245, 0.0250, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 22:11:44,464 INFO [train.py:901] (1/4) Epoch 18, batch 7250, loss[loss=0.2244, simple_loss=0.2853, pruned_loss=0.0818, over 7804.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2944, pruned_loss=0.0671, over 1614643.26 frames. ], batch size: 19, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:11:54,473 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.379e+02 2.816e+02 3.627e+02 9.857e+02, threshold=5.632e+02, percent-clipped=4.0 +2023-02-06 22:12:19,764 INFO [train.py:901] (1/4) Epoch 18, batch 7300, loss[loss=0.1703, simple_loss=0.2543, pruned_loss=0.04316, over 7793.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2936, pruned_loss=0.06632, over 1615713.48 frames. ], batch size: 19, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:12:53,982 INFO [train.py:901] (1/4) Epoch 18, batch 7350, loss[loss=0.2101, simple_loss=0.2841, pruned_loss=0.06805, over 8713.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2934, pruned_loss=0.06626, over 1621385.50 frames. ], batch size: 39, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:13:02,925 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=144773.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:13:04,749 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.484e+02 2.992e+02 3.514e+02 8.978e+02, threshold=5.985e+02, percent-clipped=6.0 +2023-02-06 22:13:08,084 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 22:13:26,891 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 22:13:28,903 INFO [train.py:901] (1/4) Epoch 18, batch 7400, loss[loss=0.2256, simple_loss=0.2992, pruned_loss=0.07604, over 7531.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2932, pruned_loss=0.06611, over 1623123.84 frames. ], batch size: 74, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:14:04,298 INFO [train.py:901] (1/4) Epoch 18, batch 7450, loss[loss=0.2603, simple_loss=0.3349, pruned_loss=0.09286, over 8425.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2917, pruned_loss=0.06528, over 1621074.17 frames. ], batch size: 48, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:14:07,790 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 22:14:14,578 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 2.433e+02 3.083e+02 4.140e+02 9.921e+02, threshold=6.167e+02, percent-clipped=3.0 +2023-02-06 22:14:23,398 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144888.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:14:30,666 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9336, 2.0513, 1.8417, 2.5900, 1.2526, 1.5355, 1.8046, 2.1016], + device='cuda:1'), covar=tensor([0.0691, 0.0807, 0.0838, 0.0382, 0.1106, 0.1311, 0.0924, 0.0704], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0197, 0.0246, 0.0210, 0.0205, 0.0244, 0.0250, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 22:14:38,536 INFO [train.py:901] (1/4) Epoch 18, batch 7500, loss[loss=0.2086, simple_loss=0.2967, pruned_loss=0.06022, over 8363.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.292, pruned_loss=0.06551, over 1622475.00 frames. ], batch size: 24, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:15:12,908 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7877, 1.4594, 2.8559, 1.3963, 2.1856, 3.0318, 3.2026, 2.6246], + device='cuda:1'), covar=tensor([0.1048, 0.1616, 0.0401, 0.2085, 0.0946, 0.0309, 0.0571, 0.0621], + device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0317, 0.0277, 0.0308, 0.0296, 0.0256, 0.0401, 0.0300], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 22:15:14,116 INFO [train.py:901] (1/4) Epoch 18, batch 7550, loss[loss=0.1615, simple_loss=0.2422, pruned_loss=0.04042, over 7430.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2924, pruned_loss=0.06568, over 1622739.94 frames. ], batch size: 17, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:15:24,764 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 2.361e+02 2.893e+02 3.293e+02 8.578e+02, threshold=5.785e+02, percent-clipped=2.0 +2023-02-06 22:15:48,829 INFO [train.py:901] (1/4) Epoch 18, batch 7600, loss[loss=0.2067, simple_loss=0.2958, pruned_loss=0.05881, over 8329.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2927, pruned_loss=0.06558, over 1623780.66 frames. ], batch size: 25, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:15:51,069 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145014.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:16:12,832 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:16:24,428 INFO [train.py:901] (1/4) Epoch 18, batch 7650, loss[loss=0.2416, simple_loss=0.3157, pruned_loss=0.08373, over 8453.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2913, pruned_loss=0.06485, over 1619678.46 frames. ], batch size: 25, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:16:35,692 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.409e+02 3.204e+02 3.806e+02 7.453e+02, threshold=6.408e+02, percent-clipped=5.0 +2023-02-06 22:16:40,595 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145084.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:16:58,568 INFO [train.py:901] (1/4) Epoch 18, batch 7700, loss[loss=0.2334, simple_loss=0.3086, pruned_loss=0.07913, over 8019.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2929, pruned_loss=0.06549, over 1623429.45 frames. ], batch size: 22, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:17:16,299 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 22:17:21,897 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145144.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:17:33,661 INFO [train.py:901] (1/4) Epoch 18, batch 7750, loss[loss=0.2037, simple_loss=0.2985, pruned_loss=0.05448, over 8249.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2928, pruned_loss=0.06567, over 1618878.89 frames. ], batch size: 24, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:17:40,034 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145169.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:17:44,036 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4628, 1.6460, 4.5572, 2.1057, 2.5697, 5.1843, 5.2521, 4.5773], + device='cuda:1'), covar=tensor([0.1097, 0.1811, 0.0240, 0.1835, 0.1124, 0.0164, 0.0261, 0.0519], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0314, 0.0276, 0.0306, 0.0294, 0.0255, 0.0399, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 22:17:45,202 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.617e+02 3.101e+02 3.765e+02 9.296e+02, threshold=6.202e+02, percent-clipped=3.0 +2023-02-06 22:18:08,805 INFO [train.py:901] (1/4) Epoch 18, batch 7800, loss[loss=0.1827, simple_loss=0.2798, pruned_loss=0.04278, over 8245.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2919, pruned_loss=0.06555, over 1614987.57 frames. ], batch size: 24, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:18:30,125 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6322, 1.5859, 2.1485, 1.5386, 1.2098, 2.0541, 0.3638, 1.3603], + device='cuda:1'), covar=tensor([0.1648, 0.1389, 0.0348, 0.1128, 0.2905, 0.0466, 0.2112, 0.1299], + device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0190, 0.0121, 0.0215, 0.0266, 0.0129, 0.0165, 0.0182], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 22:18:36,372 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6254, 1.5603, 2.0660, 1.3888, 1.2248, 2.0326, 0.3601, 1.2510], + device='cuda:1'), covar=tensor([0.1646, 0.1407, 0.0349, 0.1183, 0.2830, 0.0472, 0.2184, 0.1473], + device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0190, 0.0121, 0.0215, 0.0266, 0.0129, 0.0165, 0.0182], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 22:18:42,841 INFO [train.py:901] (1/4) Epoch 18, batch 7850, loss[loss=0.2494, simple_loss=0.3259, pruned_loss=0.08644, over 8516.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2927, pruned_loss=0.06596, over 1616880.25 frames. ], batch size: 28, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:18:53,261 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.477e+02 2.948e+02 3.643e+02 1.044e+03, threshold=5.895e+02, percent-clipped=9.0 +2023-02-06 22:19:03,116 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.34 vs. limit=5.0 +2023-02-06 22:19:16,121 INFO [train.py:901] (1/4) Epoch 18, batch 7900, loss[loss=0.2099, simple_loss=0.3046, pruned_loss=0.05758, over 8562.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2933, pruned_loss=0.0665, over 1616879.22 frames. ], batch size: 34, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:19:28,931 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5824, 1.8097, 2.0101, 1.2684, 2.0788, 1.3960, 0.5416, 1.7847], + device='cuda:1'), covar=tensor([0.0514, 0.0329, 0.0230, 0.0503, 0.0321, 0.0825, 0.0801, 0.0248], + device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0380, 0.0324, 0.0433, 0.0362, 0.0526, 0.0382, 0.0404], + device='cuda:1'), out_proj_covar=tensor([1.1925e-04, 1.0051e-04, 8.5804e-05, 1.1502e-04, 9.6359e-05, 1.5044e-04, + 1.0373e-04, 1.0803e-04], device='cuda:1') +2023-02-06 22:19:44,169 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4779, 2.3437, 3.2568, 2.4817, 3.0164, 2.4962, 2.1718, 1.7751], + device='cuda:1'), covar=tensor([0.5185, 0.5014, 0.1796, 0.3680, 0.2521, 0.2929, 0.1932, 0.5568], + device='cuda:1'), in_proj_covar=tensor([0.0939, 0.0956, 0.0783, 0.0921, 0.0988, 0.0876, 0.0737, 0.0817], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 22:19:47,150 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=145358.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:19:49,086 INFO [train.py:901] (1/4) Epoch 18, batch 7950, loss[loss=0.2348, simple_loss=0.3214, pruned_loss=0.07409, over 8478.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.293, pruned_loss=0.06606, over 1611670.59 frames. ], batch size: 29, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:19:58,001 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7958, 2.0761, 1.6279, 2.5842, 1.2196, 1.4744, 1.9038, 2.0264], + device='cuda:1'), covar=tensor([0.0771, 0.0748, 0.0931, 0.0347, 0.1077, 0.1305, 0.0815, 0.0760], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0197, 0.0248, 0.0210, 0.0206, 0.0244, 0.0250, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 22:19:59,836 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.389e+02 3.012e+02 3.869e+02 1.111e+03, threshold=6.025e+02, percent-clipped=3.0 +2023-02-06 22:20:07,941 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=145389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:20:08,011 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:20:23,114 INFO [train.py:901] (1/4) Epoch 18, batch 8000, loss[loss=0.2426, simple_loss=0.3169, pruned_loss=0.08417, over 8683.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2917, pruned_loss=0.06522, over 1614510.45 frames. ], batch size: 39, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:20:34,647 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=145428.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:20:57,108 INFO [train.py:901] (1/4) Epoch 18, batch 8050, loss[loss=0.2247, simple_loss=0.292, pruned_loss=0.0787, over 7549.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2899, pruned_loss=0.06528, over 1590885.98 frames. ], batch size: 18, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:21:05,674 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145473.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:21:08,160 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.350e+02 2.866e+02 3.408e+02 5.747e+02, threshold=5.732e+02, percent-clipped=0.0 +2023-02-06 22:21:29,502 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 22:21:34,891 INFO [train.py:901] (1/4) Epoch 19, batch 0, loss[loss=0.2417, simple_loss=0.3217, pruned_loss=0.08084, over 8700.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3217, pruned_loss=0.08084, over 8700.00 frames. ], batch size: 34, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:21:34,891 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 22:21:46,551 INFO [train.py:935] (1/4) Epoch 19, validation: loss=0.1782, simple_loss=0.2779, pruned_loss=0.03928, over 944034.00 frames. +2023-02-06 22:21:46,552 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 22:21:54,194 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:22:03,055 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 22:22:22,467 INFO [train.py:901] (1/4) Epoch 19, batch 50, loss[loss=0.1984, simple_loss=0.2856, pruned_loss=0.05564, over 8353.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2937, pruned_loss=0.06684, over 362714.64 frames. ], batch size: 24, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:22:22,664 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:22:23,341 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8206, 1.4514, 2.8276, 1.3919, 2.1287, 3.0293, 3.1765, 2.5634], + device='cuda:1'), covar=tensor([0.1025, 0.1647, 0.0444, 0.2139, 0.0996, 0.0304, 0.0643, 0.0612], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0312, 0.0275, 0.0303, 0.0293, 0.0253, 0.0396, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 22:22:25,795 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-06 22:22:40,530 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 22:22:42,317 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-06 22:22:45,196 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.340e+02 2.977e+02 3.641e+02 7.952e+02, threshold=5.953e+02, percent-clipped=6.0 +2023-02-06 22:22:56,255 INFO [train.py:901] (1/4) Epoch 19, batch 100, loss[loss=0.1941, simple_loss=0.2676, pruned_loss=0.06035, over 7305.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2949, pruned_loss=0.06676, over 640186.66 frames. ], batch size: 16, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:22:59,330 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7508, 2.2637, 4.2002, 1.5351, 3.0005, 2.2907, 1.8378, 2.8245], + device='cuda:1'), covar=tensor([0.1889, 0.2712, 0.0818, 0.4657, 0.1945, 0.3185, 0.2288, 0.2582], + device='cuda:1'), in_proj_covar=tensor([0.0515, 0.0578, 0.0548, 0.0628, 0.0635, 0.0585, 0.0517, 0.0629], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:23:01,906 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 22:23:10,101 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145612.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:23:20,693 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7284, 4.7064, 4.2140, 2.2590, 4.1561, 4.3504, 4.3300, 4.0467], + device='cuda:1'), covar=tensor([0.0679, 0.0480, 0.0880, 0.4248, 0.0748, 0.0896, 0.1049, 0.0686], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0426, 0.0429, 0.0527, 0.0414, 0.0426, 0.0408, 0.0373], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:23:32,313 INFO [train.py:901] (1/4) Epoch 19, batch 150, loss[loss=0.1784, simple_loss=0.2483, pruned_loss=0.05424, over 7689.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2952, pruned_loss=0.0669, over 859877.67 frames. ], batch size: 18, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:23:40,879 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9753, 1.7458, 2.3615, 2.0175, 2.2920, 1.9746, 1.7385, 1.2433], + device='cuda:1'), covar=tensor([0.5288, 0.4764, 0.1702, 0.3045, 0.2156, 0.2955, 0.2062, 0.4657], + device='cuda:1'), in_proj_covar=tensor([0.0925, 0.0945, 0.0773, 0.0911, 0.0976, 0.0864, 0.0728, 0.0808], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 22:23:46,188 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:23:57,038 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.454e+02 2.969e+02 3.777e+02 1.176e+03, threshold=5.938e+02, percent-clipped=4.0 +2023-02-06 22:24:07,984 INFO [train.py:901] (1/4) Epoch 19, batch 200, loss[loss=0.2118, simple_loss=0.2929, pruned_loss=0.06531, over 8344.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2944, pruned_loss=0.06658, over 1027778.06 frames. ], batch size: 26, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:24:33,099 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4630, 2.5273, 1.8006, 2.2031, 2.0766, 1.4971, 2.0719, 1.9556], + device='cuda:1'), covar=tensor([0.1488, 0.0392, 0.1090, 0.0553, 0.0663, 0.1400, 0.0839, 0.0965], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0234, 0.0323, 0.0301, 0.0296, 0.0328, 0.0339, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 22:24:33,115 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145729.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:24:35,725 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=145733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:24:43,207 INFO [train.py:901] (1/4) Epoch 19, batch 250, loss[loss=0.1688, simple_loss=0.2505, pruned_loss=0.04355, over 7689.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2936, pruned_loss=0.06552, over 1164728.25 frames. ], batch size: 18, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:24:51,124 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:24:55,239 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:24:58,385 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 22:25:06,966 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 22:25:07,541 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 2.432e+02 3.022e+02 3.893e+02 7.688e+02, threshold=6.043e+02, percent-clipped=6.0 +2023-02-06 22:25:13,282 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:25:18,657 INFO [train.py:901] (1/4) Epoch 19, batch 300, loss[loss=0.2015, simple_loss=0.2747, pruned_loss=0.06417, over 5620.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2944, pruned_loss=0.06619, over 1261555.60 frames. ], batch size: 12, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:25:22,936 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145799.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:25:39,979 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:25:50,255 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145839.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:25:53,700 INFO [train.py:901] (1/4) Epoch 19, batch 350, loss[loss=0.1896, simple_loss=0.2855, pruned_loss=0.04688, over 8096.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2951, pruned_loss=0.06651, over 1343511.58 frames. ], batch size: 23, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:25:57,437 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:26:17,675 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.383e+02 2.952e+02 3.795e+02 9.100e+02, threshold=5.904e+02, percent-clipped=6.0 +2023-02-06 22:26:30,025 INFO [train.py:901] (1/4) Epoch 19, batch 400, loss[loss=0.1615, simple_loss=0.2393, pruned_loss=0.0418, over 7529.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2942, pruned_loss=0.06555, over 1407846.24 frames. ], batch size: 18, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:26:32,906 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4578, 4.4775, 4.0250, 1.9145, 3.9783, 4.0936, 3.9563, 3.8319], + device='cuda:1'), covar=tensor([0.0773, 0.0525, 0.1162, 0.5217, 0.0873, 0.0997, 0.1273, 0.0810], + device='cuda:1'), in_proj_covar=tensor([0.0511, 0.0422, 0.0425, 0.0524, 0.0411, 0.0424, 0.0402, 0.0370], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:26:44,748 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2697, 1.5873, 1.7973, 1.4441, 1.2097, 1.6461, 2.0862, 1.9880], + device='cuda:1'), covar=tensor([0.0526, 0.1245, 0.1617, 0.1456, 0.0616, 0.1423, 0.0650, 0.0556], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0157, 0.0099, 0.0161, 0.0112, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 22:27:04,035 INFO [train.py:901] (1/4) Epoch 19, batch 450, loss[loss=0.1679, simple_loss=0.2507, pruned_loss=0.04252, over 8365.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2957, pruned_loss=0.06678, over 1456966.19 frames. ], batch size: 24, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:27:12,891 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=145956.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:27:23,357 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.20 vs. limit=5.0 +2023-02-06 22:27:28,526 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.474e+02 2.839e+02 3.457e+02 5.406e+02, threshold=5.677e+02, percent-clipped=0.0 +2023-02-06 22:27:40,183 INFO [train.py:901] (1/4) Epoch 19, batch 500, loss[loss=0.1895, simple_loss=0.2632, pruned_loss=0.05786, over 7982.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2935, pruned_loss=0.06613, over 1488719.93 frames. ], batch size: 21, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:27:50,094 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:28:03,401 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5710, 1.3915, 4.7434, 1.7784, 4.2137, 3.8954, 4.2972, 4.1226], + device='cuda:1'), covar=tensor([0.0570, 0.4932, 0.0409, 0.3971, 0.0998, 0.0926, 0.0561, 0.0646], + device='cuda:1'), in_proj_covar=tensor([0.0597, 0.0627, 0.0673, 0.0605, 0.0687, 0.0591, 0.0585, 0.0649], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:28:15,841 INFO [train.py:901] (1/4) Epoch 19, batch 550, loss[loss=0.2071, simple_loss=0.2898, pruned_loss=0.06219, over 8524.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2923, pruned_loss=0.06513, over 1516839.15 frames. ], batch size: 39, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:28:19,297 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7064, 4.7183, 4.2677, 1.8156, 4.2075, 4.3123, 4.3906, 4.0287], + device='cuda:1'), covar=tensor([0.0625, 0.0468, 0.0941, 0.4593, 0.0743, 0.0916, 0.1053, 0.0710], + device='cuda:1'), in_proj_covar=tensor([0.0519, 0.0428, 0.0430, 0.0530, 0.0416, 0.0430, 0.0411, 0.0375], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:28:35,104 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146071.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:28:38,934 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.557e+02 3.049e+02 4.000e+02 8.642e+02, threshold=6.099e+02, percent-clipped=4.0 +2023-02-06 22:28:50,773 INFO [train.py:901] (1/4) Epoch 19, batch 600, loss[loss=0.1963, simple_loss=0.2802, pruned_loss=0.05621, over 8658.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2939, pruned_loss=0.06588, over 1537246.40 frames. ], batch size: 34, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:28:54,549 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4431, 1.2818, 2.5470, 0.9591, 2.2805, 2.1384, 2.3258, 2.2776], + device='cuda:1'), covar=tensor([0.0737, 0.2930, 0.0986, 0.3432, 0.1142, 0.1080, 0.0670, 0.0757], + device='cuda:1'), in_proj_covar=tensor([0.0594, 0.0624, 0.0670, 0.0600, 0.0684, 0.0588, 0.0582, 0.0646], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 22:28:59,454 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146104.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:29:03,469 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3733, 4.2899, 3.9130, 2.0367, 3.8768, 3.9815, 3.9809, 3.7188], + device='cuda:1'), covar=tensor([0.0770, 0.0563, 0.1035, 0.4750, 0.0816, 0.1071, 0.1260, 0.0812], + device='cuda:1'), in_proj_covar=tensor([0.0517, 0.0425, 0.0427, 0.0526, 0.0413, 0.0427, 0.0407, 0.0373], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:29:11,428 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 22:29:11,608 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:29:17,642 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146129.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:29:26,663 INFO [train.py:901] (1/4) Epoch 19, batch 650, loss[loss=0.2183, simple_loss=0.3046, pruned_loss=0.06599, over 8021.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2938, pruned_loss=0.06569, over 1558214.07 frames. ], batch size: 22, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:29:42,927 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146167.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:29:49,753 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 2.628e+02 2.995e+02 3.912e+02 8.872e+02, threshold=5.991e+02, percent-clipped=7.0 +2023-02-06 22:29:53,934 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146183.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:30:00,618 INFO [train.py:901] (1/4) Epoch 19, batch 700, loss[loss=0.2443, simple_loss=0.3182, pruned_loss=0.08524, over 8328.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.295, pruned_loss=0.06642, over 1572025.73 frames. ], batch size: 25, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:30:37,723 INFO [train.py:901] (1/4) Epoch 19, batch 750, loss[loss=0.239, simple_loss=0.3227, pruned_loss=0.07767, over 8495.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2937, pruned_loss=0.0654, over 1586531.66 frames. ], batch size: 26, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:30:58,076 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 22:31:00,739 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.187e+02 2.733e+02 3.387e+02 1.037e+03, threshold=5.466e+02, percent-clipped=4.0 +2023-02-06 22:31:06,863 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 22:31:11,486 INFO [train.py:901] (1/4) Epoch 19, batch 800, loss[loss=0.2092, simple_loss=0.2839, pruned_loss=0.06723, over 7964.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2943, pruned_loss=0.06611, over 1594955.29 frames. ], batch size: 21, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:31:14,847 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146298.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:31:15,519 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0349, 1.4527, 1.6441, 1.3550, 0.9987, 1.4658, 1.7496, 1.5376], + device='cuda:1'), covar=tensor([0.0487, 0.1295, 0.1717, 0.1479, 0.0588, 0.1535, 0.0677, 0.0696], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0158, 0.0099, 0.0162, 0.0113, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 22:31:17,172 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.41 vs. limit=5.0 +2023-02-06 22:31:35,743 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:31:47,234 INFO [train.py:901] (1/4) Epoch 19, batch 850, loss[loss=0.2036, simple_loss=0.2916, pruned_loss=0.0578, over 8290.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.294, pruned_loss=0.0659, over 1601380.67 frames. ], batch size: 23, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:31:54,226 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146352.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:32:10,846 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:32:11,292 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.470e+02 3.071e+02 3.941e+02 1.675e+03, threshold=6.141e+02, percent-clipped=6.0 +2023-02-06 22:32:22,254 INFO [train.py:901] (1/4) Epoch 19, batch 900, loss[loss=0.1937, simple_loss=0.2728, pruned_loss=0.05733, over 8515.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.294, pruned_loss=0.06618, over 1601572.16 frames. ], batch size: 28, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:32:27,826 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146401.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:32:41,200 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8208, 2.2532, 4.0353, 1.5560, 2.9932, 2.1967, 1.8680, 2.6409], + device='cuda:1'), covar=tensor([0.1646, 0.2227, 0.0699, 0.3916, 0.1467, 0.2816, 0.1913, 0.2244], + device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0580, 0.0550, 0.0630, 0.0639, 0.0584, 0.0518, 0.0629], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:32:56,365 INFO [train.py:901] (1/4) Epoch 19, batch 950, loss[loss=0.2148, simple_loss=0.2916, pruned_loss=0.06897, over 8628.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2948, pruned_loss=0.06634, over 1607568.80 frames. ], batch size: 34, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:33:09,722 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.47 vs. limit=5.0 +2023-02-06 22:33:20,835 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1515, 1.3087, 1.5822, 1.2088, 0.9145, 1.3080, 1.5737, 1.5461], + device='cuda:1'), covar=tensor([0.0479, 0.1313, 0.1670, 0.1499, 0.0621, 0.1556, 0.0706, 0.0652], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0188, 0.0157, 0.0099, 0.0162, 0.0112, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 22:33:21,316 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 2.324e+02 2.987e+02 4.077e+02 9.877e+02, threshold=5.974e+02, percent-clipped=4.0 +2023-02-06 22:33:22,714 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 22:33:32,154 INFO [train.py:901] (1/4) Epoch 19, batch 1000, loss[loss=0.2414, simple_loss=0.3187, pruned_loss=0.08207, over 8410.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2938, pruned_loss=0.06574, over 1608029.56 frames. ], batch size: 48, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:33:44,466 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146511.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:33:54,570 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 22:34:06,365 INFO [train.py:901] (1/4) Epoch 19, batch 1050, loss[loss=0.219, simple_loss=0.3016, pruned_loss=0.0682, over 8364.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.294, pruned_loss=0.06559, over 1607523.31 frames. ], batch size: 24, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:34:06,379 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 22:34:14,963 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146554.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:34:31,585 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 2.403e+02 2.837e+02 3.508e+02 6.242e+02, threshold=5.674e+02, percent-clipped=1.0 +2023-02-06 22:34:32,762 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-02-06 22:34:33,867 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146579.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:34:44,087 INFO [train.py:901] (1/4) Epoch 19, batch 1100, loss[loss=0.1983, simple_loss=0.2897, pruned_loss=0.05342, over 8361.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2923, pruned_loss=0.06535, over 1607312.86 frames. ], batch size: 24, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:34:55,173 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146609.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:35:06,980 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146626.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:35:18,380 INFO [train.py:901] (1/4) Epoch 19, batch 1150, loss[loss=0.2098, simple_loss=0.2781, pruned_loss=0.07071, over 7430.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2918, pruned_loss=0.065, over 1609797.37 frames. ], batch size: 17, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:35:19,122 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 22:35:19,277 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:35:42,416 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 2.484e+02 2.879e+02 3.755e+02 5.922e+02, threshold=5.758e+02, percent-clipped=3.0 +2023-02-06 22:35:53,863 INFO [train.py:901] (1/4) Epoch 19, batch 1200, loss[loss=0.2106, simple_loss=0.2838, pruned_loss=0.06869, over 7814.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2923, pruned_loss=0.06552, over 1614730.46 frames. ], batch size: 20, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:36:24,675 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9262, 1.5484, 3.2936, 1.4576, 2.1852, 3.5613, 3.6460, 3.0279], + device='cuda:1'), covar=tensor([0.1093, 0.1660, 0.0335, 0.2071, 0.1094, 0.0240, 0.0655, 0.0550], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0314, 0.0280, 0.0307, 0.0297, 0.0256, 0.0400, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 22:36:29,002 INFO [train.py:901] (1/4) Epoch 19, batch 1250, loss[loss=0.1808, simple_loss=0.2569, pruned_loss=0.0524, over 7701.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2932, pruned_loss=0.06594, over 1614225.89 frames. ], batch size: 18, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:36:34,072 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2891, 1.9113, 2.5680, 2.0959, 2.4600, 2.2657, 2.0222, 1.3466], + device='cuda:1'), covar=tensor([0.5141, 0.4809, 0.1893, 0.3666, 0.2337, 0.2824, 0.1930, 0.5041], + device='cuda:1'), in_proj_covar=tensor([0.0929, 0.0947, 0.0781, 0.0912, 0.0978, 0.0864, 0.0726, 0.0807], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 22:36:52,652 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 2.471e+02 2.976e+02 4.092e+02 7.603e+02, threshold=5.951e+02, percent-clipped=4.0 +2023-02-06 22:37:04,305 INFO [train.py:901] (1/4) Epoch 19, batch 1300, loss[loss=0.1904, simple_loss=0.278, pruned_loss=0.05134, over 8375.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.293, pruned_loss=0.06556, over 1614747.18 frames. ], batch size: 24, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:37:40,719 INFO [train.py:901] (1/4) Epoch 19, batch 1350, loss[loss=0.1972, simple_loss=0.2821, pruned_loss=0.05613, over 8081.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2934, pruned_loss=0.06552, over 1615302.56 frames. ], batch size: 21, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:37:53,728 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146862.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:37:58,975 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-06 22:38:03,894 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.302e+02 2.844e+02 3.659e+02 6.626e+02, threshold=5.688e+02, percent-clipped=1.0 +2023-02-06 22:38:07,775 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146882.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:38:15,229 INFO [train.py:901] (1/4) Epoch 19, batch 1400, loss[loss=0.1871, simple_loss=0.2737, pruned_loss=0.05029, over 7810.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2918, pruned_loss=0.06476, over 1612316.36 frames. ], batch size: 20, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:38:25,961 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146907.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:38:43,245 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 22:38:52,627 INFO [train.py:901] (1/4) Epoch 19, batch 1450, loss[loss=0.2365, simple_loss=0.3205, pruned_loss=0.07621, over 8243.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2924, pruned_loss=0.06489, over 1616857.69 frames. ], batch size: 24, lr: 4.03e-03, grad_scale: 16.0 +2023-02-06 22:38:56,691 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 22:38:59,396 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146953.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:39:16,186 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.810e+02 2.362e+02 2.962e+02 3.993e+02 1.525e+03, threshold=5.923e+02, percent-clipped=6.0 +2023-02-06 22:39:22,647 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6057, 1.3274, 4.8696, 1.7794, 4.2414, 4.0737, 4.3485, 4.2639], + device='cuda:1'), covar=tensor([0.0702, 0.5073, 0.0574, 0.4227, 0.1277, 0.1038, 0.0705, 0.0713], + device='cuda:1'), in_proj_covar=tensor([0.0602, 0.0635, 0.0677, 0.0609, 0.0692, 0.0594, 0.0590, 0.0651], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:39:23,969 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146988.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:39:27,292 INFO [train.py:901] (1/4) Epoch 19, batch 1500, loss[loss=0.1892, simple_loss=0.2632, pruned_loss=0.05755, over 7437.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2912, pruned_loss=0.0644, over 1613930.35 frames. ], batch size: 17, lr: 4.03e-03, grad_scale: 16.0 +2023-02-06 22:40:03,272 INFO [train.py:901] (1/4) Epoch 19, batch 1550, loss[loss=0.1913, simple_loss=0.2856, pruned_loss=0.04853, over 8104.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2913, pruned_loss=0.06395, over 1614169.51 frames. ], batch size: 23, lr: 4.03e-03, grad_scale: 16.0 +2023-02-06 22:40:22,634 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147068.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:40:28,388 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.434e+02 2.984e+02 3.600e+02 8.495e+02, threshold=5.968e+02, percent-clipped=1.0 +2023-02-06 22:40:39,458 INFO [train.py:901] (1/4) Epoch 19, batch 1600, loss[loss=0.1683, simple_loss=0.2568, pruned_loss=0.03995, over 7542.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2916, pruned_loss=0.06378, over 1617136.95 frames. ], batch size: 18, lr: 4.03e-03, grad_scale: 16.0 +2023-02-06 22:40:46,387 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147103.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:41:14,549 INFO [train.py:901] (1/4) Epoch 19, batch 1650, loss[loss=0.1876, simple_loss=0.2618, pruned_loss=0.05674, over 7444.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2911, pruned_loss=0.06412, over 1614394.24 frames. ], batch size: 17, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:41:40,971 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.354e+02 2.709e+02 3.474e+02 7.081e+02, threshold=5.418e+02, percent-clipped=1.0 +2023-02-06 22:41:51,243 INFO [train.py:901] (1/4) Epoch 19, batch 1700, loss[loss=0.2374, simple_loss=0.3119, pruned_loss=0.08147, over 8582.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2901, pruned_loss=0.06369, over 1616118.35 frames. ], batch size: 34, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:41:52,266 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0857, 1.8117, 2.3475, 1.9647, 2.3053, 2.1213, 1.8884, 1.1316], + device='cuda:1'), covar=tensor([0.5138, 0.4573, 0.1816, 0.3609, 0.2403, 0.2910, 0.1818, 0.4881], + device='cuda:1'), in_proj_covar=tensor([0.0936, 0.0953, 0.0788, 0.0917, 0.0984, 0.0869, 0.0730, 0.0810], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 22:42:00,574 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=147206.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:42:25,938 INFO [train.py:901] (1/4) Epoch 19, batch 1750, loss[loss=0.2057, simple_loss=0.2881, pruned_loss=0.06163, over 8318.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2903, pruned_loss=0.06337, over 1614595.14 frames. ], batch size: 25, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:42:37,957 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147259.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:42:41,406 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3600, 1.5899, 4.5623, 1.6819, 3.9810, 3.7949, 4.0827, 3.9685], + device='cuda:1'), covar=tensor([0.0651, 0.4741, 0.0518, 0.4351, 0.1167, 0.1032, 0.0658, 0.0704], + device='cuda:1'), in_proj_covar=tensor([0.0601, 0.0636, 0.0677, 0.0612, 0.0692, 0.0596, 0.0591, 0.0653], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:42:46,385 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3957, 4.4194, 3.9720, 2.1516, 3.8962, 4.0056, 3.9524, 3.7968], + device='cuda:1'), covar=tensor([0.0705, 0.0509, 0.0955, 0.4273, 0.0810, 0.1019, 0.1348, 0.0785], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0423, 0.0428, 0.0526, 0.0415, 0.0427, 0.0410, 0.0375], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:42:51,049 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.541e+02 2.529e+02 3.043e+02 3.569e+02 7.736e+02, threshold=6.085e+02, percent-clipped=5.0 +2023-02-06 22:43:03,013 INFO [train.py:901] (1/4) Epoch 19, batch 1800, loss[loss=0.1835, simple_loss=0.2582, pruned_loss=0.0544, over 7705.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2916, pruned_loss=0.06431, over 1614899.50 frames. ], batch size: 18, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:43:22,536 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147321.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:43:24,599 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147324.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:43:27,343 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:43:27,657 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 22:43:37,466 INFO [train.py:901] (1/4) Epoch 19, batch 1850, loss[loss=0.2229, simple_loss=0.3118, pruned_loss=0.06694, over 8564.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2917, pruned_loss=0.06445, over 1616607.39 frames. ], batch size: 34, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:43:41,699 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:43:48,548 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:44:02,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.300e+02 2.823e+02 3.606e+02 1.006e+03, threshold=5.645e+02, percent-clipped=2.0 +2023-02-06 22:44:02,738 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 22:44:06,650 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147384.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:44:12,523 INFO [train.py:901] (1/4) Epoch 19, batch 1900, loss[loss=0.1965, simple_loss=0.2851, pruned_loss=0.05393, over 8331.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2906, pruned_loss=0.06418, over 1612564.32 frames. ], batch size: 26, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:44:37,275 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147425.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:44:44,950 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 22:44:49,667 INFO [train.py:901] (1/4) Epoch 19, batch 1950, loss[loss=0.256, simple_loss=0.3314, pruned_loss=0.09027, over 8030.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2905, pruned_loss=0.06438, over 1607730.47 frames. ], batch size: 22, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:44:55,957 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147452.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:44:56,507 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 22:45:13,742 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.289e+02 2.862e+02 3.830e+02 8.439e+02, threshold=5.724e+02, percent-clipped=6.0 +2023-02-06 22:45:15,255 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 22:45:24,879 INFO [train.py:901] (1/4) Epoch 19, batch 2000, loss[loss=0.2109, simple_loss=0.2871, pruned_loss=0.0673, over 7822.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2917, pruned_loss=0.06499, over 1612180.38 frames. ], batch size: 20, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:46:01,778 INFO [train.py:901] (1/4) Epoch 19, batch 2050, loss[loss=0.1868, simple_loss=0.2848, pruned_loss=0.04437, over 8334.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2902, pruned_loss=0.06495, over 1604939.24 frames. ], batch size: 26, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:46:09,084 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.02 vs. limit=5.0 +2023-02-06 22:46:25,331 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147577.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:46:25,778 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.500e+02 2.918e+02 3.445e+02 6.516e+02, threshold=5.836e+02, percent-clipped=2.0 +2023-02-06 22:46:36,262 INFO [train.py:901] (1/4) Epoch 19, batch 2100, loss[loss=0.2038, simple_loss=0.2949, pruned_loss=0.05641, over 8301.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2902, pruned_loss=0.06487, over 1606888.50 frames. ], batch size: 23, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:46:42,944 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147602.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:46:43,436 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=147603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:46:47,318 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.69 vs. limit=5.0 +2023-02-06 22:47:12,101 INFO [train.py:901] (1/4) Epoch 19, batch 2150, loss[loss=0.2346, simple_loss=0.3225, pruned_loss=0.07341, over 8514.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2899, pruned_loss=0.06461, over 1612279.31 frames. ], batch size: 26, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:47:28,882 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0185, 3.5909, 2.0898, 2.5490, 2.6623, 2.0042, 2.6232, 2.8096], + device='cuda:1'), covar=tensor([0.1668, 0.0330, 0.1172, 0.0879, 0.0846, 0.1357, 0.1040, 0.1073], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0234, 0.0325, 0.0303, 0.0301, 0.0331, 0.0341, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 22:47:31,900 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-06 22:47:33,545 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=147672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:47:36,908 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147677.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:47:37,423 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.398e+02 3.174e+02 3.852e+02 9.466e+02, threshold=6.348e+02, percent-clipped=6.0 +2023-02-06 22:47:47,724 INFO [train.py:901] (1/4) Epoch 19, batch 2200, loss[loss=0.1917, simple_loss=0.2829, pruned_loss=0.05028, over 8202.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2906, pruned_loss=0.06481, over 1614805.61 frames. ], batch size: 23, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:48:04,720 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:48:21,943 INFO [train.py:901] (1/4) Epoch 19, batch 2250, loss[loss=0.1978, simple_loss=0.2869, pruned_loss=0.05431, over 8530.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2915, pruned_loss=0.06544, over 1614737.54 frames. ], batch size: 39, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:48:41,107 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=147769.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:48:47,103 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.392e+02 3.089e+02 3.849e+02 9.613e+02, threshold=6.179e+02, percent-clipped=2.0 +2023-02-06 22:48:53,334 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147787.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:48:56,992 INFO [train.py:901] (1/4) Epoch 19, batch 2300, loss[loss=0.2157, simple_loss=0.2894, pruned_loss=0.07099, over 8137.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2911, pruned_loss=0.06499, over 1614251.06 frames. ], batch size: 22, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:48:58,973 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=147796.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:49:24,342 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:49:30,996 INFO [train.py:901] (1/4) Epoch 19, batch 2350, loss[loss=0.2197, simple_loss=0.3061, pruned_loss=0.06662, over 8519.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2909, pruned_loss=0.06461, over 1614309.88 frames. ], batch size: 28, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:49:53,251 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:49:55,896 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.451e+02 2.984e+02 3.607e+02 1.132e+03, threshold=5.968e+02, percent-clipped=4.0 +2023-02-06 22:50:01,541 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147884.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:50:05,004 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0967, 2.2688, 1.9712, 2.8831, 1.4514, 1.7242, 2.1709, 2.3893], + device='cuda:1'), covar=tensor([0.0675, 0.0838, 0.0855, 0.0320, 0.1112, 0.1233, 0.0836, 0.0643], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0197, 0.0249, 0.0212, 0.0207, 0.0246, 0.0253, 0.0212], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 22:50:07,522 INFO [train.py:901] (1/4) Epoch 19, batch 2400, loss[loss=0.1992, simple_loss=0.2846, pruned_loss=0.05688, over 8239.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2911, pruned_loss=0.06405, over 1620235.72 frames. ], batch size: 22, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:50:17,624 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.63 vs. limit=5.0 +2023-02-06 22:50:20,048 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:50:32,259 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147929.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:50:41,460 INFO [train.py:901] (1/4) Epoch 19, batch 2450, loss[loss=0.2023, simple_loss=0.2854, pruned_loss=0.05958, over 7658.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2918, pruned_loss=0.0648, over 1616783.13 frames. ], batch size: 19, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:51:03,066 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147974.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:51:05,437 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.749e+02 2.599e+02 2.990e+02 3.557e+02 6.406e+02, threshold=5.981e+02, percent-clipped=1.0 +2023-02-06 22:51:15,597 INFO [train.py:901] (1/4) Epoch 19, batch 2500, loss[loss=0.2271, simple_loss=0.3057, pruned_loss=0.07428, over 8191.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2914, pruned_loss=0.06526, over 1615662.21 frames. ], batch size: 23, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:51:20,624 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:51:37,735 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148021.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:51:41,235 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4644, 2.0120, 3.2213, 1.2723, 2.5431, 1.9345, 1.6404, 2.4261], + device='cuda:1'), covar=tensor([0.2100, 0.2487, 0.0862, 0.4721, 0.1857, 0.3362, 0.2293, 0.2403], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0577, 0.0548, 0.0626, 0.0634, 0.0583, 0.0518, 0.0627], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 22:51:52,322 INFO [train.py:901] (1/4) Epoch 19, batch 2550, loss[loss=0.2216, simple_loss=0.298, pruned_loss=0.07261, over 8284.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2921, pruned_loss=0.06576, over 1612994.31 frames. ], batch size: 23, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:51:52,574 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:52:09,285 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148068.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:52:15,626 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.379e+02 2.867e+02 3.516e+02 7.047e+02, threshold=5.734e+02, percent-clipped=3.0 +2023-02-06 22:52:17,126 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5261, 1.8393, 2.0660, 1.2738, 2.0598, 1.3447, 0.6310, 1.7748], + device='cuda:1'), covar=tensor([0.0667, 0.0369, 0.0241, 0.0610, 0.0431, 0.0895, 0.0826, 0.0296], + device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0380, 0.0329, 0.0435, 0.0363, 0.0526, 0.0383, 0.0404], + device='cuda:1'), out_proj_covar=tensor([1.1934e-04, 1.0026e-04, 8.7018e-05, 1.1568e-04, 9.6228e-05, 1.5023e-04, + 1.0354e-04, 1.0795e-04], device='cuda:1') +2023-02-06 22:52:24,513 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0911, 1.6326, 1.3846, 1.6109, 1.3335, 1.2174, 1.3230, 1.3967], + device='cuda:1'), covar=tensor([0.0991, 0.0461, 0.1249, 0.0494, 0.0812, 0.1486, 0.0865, 0.0712], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0236, 0.0326, 0.0305, 0.0302, 0.0333, 0.0343, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 22:52:26,280 INFO [train.py:901] (1/4) Epoch 19, batch 2600, loss[loss=0.2183, simple_loss=0.2994, pruned_loss=0.06864, over 8361.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2923, pruned_loss=0.06555, over 1616077.82 frames. ], batch size: 24, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:52:57,239 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148136.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:53:00,041 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:53:01,854 INFO [train.py:901] (1/4) Epoch 19, batch 2650, loss[loss=0.2035, simple_loss=0.2887, pruned_loss=0.05918, over 8466.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2933, pruned_loss=0.06589, over 1616222.44 frames. ], batch size: 25, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:53:16,796 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:53:18,192 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148167.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:53:24,873 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148177.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:53:25,454 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 2.384e+02 2.853e+02 3.529e+02 7.126e+02, threshold=5.707e+02, percent-clipped=4.0 +2023-02-06 22:53:27,178 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 +2023-02-06 22:53:35,175 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148192.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:53:35,670 INFO [train.py:901] (1/4) Epoch 19, batch 2700, loss[loss=0.1607, simple_loss=0.2417, pruned_loss=0.03982, over 7544.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2917, pruned_loss=0.06527, over 1608690.19 frames. ], batch size: 18, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:53:38,569 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5759, 2.7901, 1.9139, 2.3653, 2.3006, 1.7128, 2.2310, 2.2127], + device='cuda:1'), covar=tensor([0.1541, 0.0384, 0.1143, 0.0640, 0.0711, 0.1377, 0.0935, 0.0999], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0237, 0.0327, 0.0306, 0.0302, 0.0333, 0.0344, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 22:53:54,170 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148219.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:54:03,913 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8748, 1.9624, 1.7071, 2.3211, 1.1250, 1.5631, 1.7236, 1.9154], + device='cuda:1'), covar=tensor([0.0661, 0.0704, 0.0950, 0.0398, 0.1063, 0.1334, 0.0802, 0.0702], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0198, 0.0252, 0.0214, 0.0208, 0.0248, 0.0256, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 22:54:11,924 INFO [train.py:901] (1/4) Epoch 19, batch 2750, loss[loss=0.232, simple_loss=0.3078, pruned_loss=0.07813, over 8185.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2909, pruned_loss=0.06486, over 1609506.74 frames. ], batch size: 23, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:54:32,993 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148273.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:54:36,065 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.455e+02 2.484e+02 2.895e+02 4.098e+02 9.310e+02, threshold=5.790e+02, percent-clipped=8.0 +2023-02-06 22:54:45,411 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148292.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:54:45,920 INFO [train.py:901] (1/4) Epoch 19, batch 2800, loss[loss=0.1951, simple_loss=0.282, pruned_loss=0.05408, over 8243.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2909, pruned_loss=0.06501, over 1614452.48 frames. ], batch size: 22, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:54:54,049 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148305.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:55:13,969 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148334.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:55:19,833 INFO [train.py:901] (1/4) Epoch 19, batch 2850, loss[loss=0.2912, simple_loss=0.336, pruned_loss=0.1232, over 7981.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2912, pruned_loss=0.06559, over 1612429.74 frames. ], batch size: 21, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:55:29,882 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.90 vs. limit=5.0 +2023-02-06 22:55:46,065 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.455e+02 2.512e+02 2.931e+02 3.824e+02 7.566e+02, threshold=5.862e+02, percent-clipped=4.0 +2023-02-06 22:55:52,986 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:55:55,652 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:55:56,116 INFO [train.py:901] (1/4) Epoch 19, batch 2900, loss[loss=0.2649, simple_loss=0.3339, pruned_loss=0.09796, over 7008.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2897, pruned_loss=0.06438, over 1606786.98 frames. ], batch size: 71, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:56:10,226 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 22:56:12,658 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:56:29,358 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 22:56:29,949 INFO [train.py:901] (1/4) Epoch 19, batch 2950, loss[loss=0.2387, simple_loss=0.3178, pruned_loss=0.07978, over 8554.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2903, pruned_loss=0.06514, over 1605597.90 frames. ], batch size: 34, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:56:32,825 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148447.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:56:52,428 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0194, 1.5289, 1.6399, 1.3652, 1.0242, 1.4514, 1.7474, 1.5306], + device='cuda:1'), covar=tensor([0.0514, 0.1285, 0.1764, 0.1483, 0.0590, 0.1557, 0.0695, 0.0670], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0158, 0.0100, 0.0162, 0.0113, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 22:56:54,932 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.514e+02 3.009e+02 3.973e+02 7.443e+02, threshold=6.017e+02, percent-clipped=3.0 +2023-02-06 22:57:06,341 INFO [train.py:901] (1/4) Epoch 19, batch 3000, loss[loss=0.2486, simple_loss=0.3345, pruned_loss=0.08139, over 8325.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2925, pruned_loss=0.06588, over 1607311.53 frames. ], batch size: 25, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:57:06,342 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 22:57:22,672 INFO [train.py:935] (1/4) Epoch 19, validation: loss=0.1752, simple_loss=0.2756, pruned_loss=0.03738, over 944034.00 frames. +2023-02-06 22:57:22,673 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 22:57:36,057 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.81 vs. limit=5.0 +2023-02-06 22:57:38,590 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148516.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:57:56,934 INFO [train.py:901] (1/4) Epoch 19, batch 3050, loss[loss=0.2342, simple_loss=0.2961, pruned_loss=0.08619, over 8234.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2931, pruned_loss=0.06605, over 1611911.30 frames. ], batch size: 22, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:58:00,751 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148548.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:58:17,724 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:58:21,052 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.398e+02 2.811e+02 3.727e+02 6.995e+02, threshold=5.622e+02, percent-clipped=3.0 +2023-02-06 22:58:30,254 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148590.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:58:32,165 INFO [train.py:901] (1/4) Epoch 19, batch 3100, loss[loss=0.2279, simple_loss=0.3134, pruned_loss=0.07117, over 8446.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2922, pruned_loss=0.06502, over 1608637.74 frames. ], batch size: 27, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:58:49,347 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:59:09,452 INFO [train.py:901] (1/4) Epoch 19, batch 3150, loss[loss=0.2476, simple_loss=0.3241, pruned_loss=0.08559, over 8560.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2918, pruned_loss=0.0647, over 1603364.54 frames. ], batch size: 31, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:59:10,338 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:59:13,422 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:59:26,361 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148669.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:59:26,422 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148669.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:59:32,308 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.732e+02 2.358e+02 3.073e+02 3.824e+02 9.523e+02, threshold=6.146e+02, percent-clipped=8.0 +2023-02-06 22:59:42,402 INFO [train.py:901] (1/4) Epoch 19, batch 3200, loss[loss=0.2212, simple_loss=0.3056, pruned_loss=0.0684, over 8556.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2928, pruned_loss=0.06538, over 1607965.82 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:00:12,967 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148734.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:00:19,662 INFO [train.py:901] (1/4) Epoch 19, batch 3250, loss[loss=0.228, simple_loss=0.3131, pruned_loss=0.07144, over 8589.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2914, pruned_loss=0.06487, over 1607930.29 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:00:34,077 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148764.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:00:43,253 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.443e+02 3.073e+02 4.112e+02 8.183e+02, threshold=6.146e+02, percent-clipped=4.0 +2023-02-06 23:00:52,317 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148791.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:00:53,595 INFO [train.py:901] (1/4) Epoch 19, batch 3300, loss[loss=0.2837, simple_loss=0.3552, pruned_loss=0.1061, over 6837.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2903, pruned_loss=0.06424, over 1604570.36 frames. ], batch size: 72, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:01:14,546 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 23:01:28,247 INFO [train.py:901] (1/4) Epoch 19, batch 3350, loss[loss=0.227, simple_loss=0.3097, pruned_loss=0.07218, over 8102.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2905, pruned_loss=0.06429, over 1607729.81 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:01:33,141 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8684, 1.4190, 2.7977, 1.3116, 2.0741, 2.9737, 3.1272, 2.3946], + device='cuda:1'), covar=tensor([0.1063, 0.1802, 0.0494, 0.2255, 0.1023, 0.0403, 0.0661, 0.0881], + device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0318, 0.0285, 0.0311, 0.0302, 0.0262, 0.0405, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-06 23:01:41,954 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148860.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:01:49,408 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7489, 2.0900, 2.2892, 1.5068, 2.2231, 1.6397, 0.7391, 2.0503], + device='cuda:1'), covar=tensor([0.0588, 0.0346, 0.0251, 0.0548, 0.0417, 0.0775, 0.0769, 0.0290], + device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0378, 0.0327, 0.0435, 0.0362, 0.0527, 0.0381, 0.0404], + device='cuda:1'), out_proj_covar=tensor([1.1907e-04, 9.9730e-05, 8.6526e-05, 1.1550e-04, 9.6168e-05, 1.5065e-04, + 1.0315e-04, 1.0799e-04], device='cuda:1') +2023-02-06 23:01:53,951 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.291e+02 2.864e+02 3.449e+02 6.722e+02, threshold=5.728e+02, percent-clipped=1.0 +2023-02-06 23:02:04,175 INFO [train.py:901] (1/4) Epoch 19, batch 3400, loss[loss=0.1915, simple_loss=0.2743, pruned_loss=0.05437, over 8115.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2904, pruned_loss=0.0644, over 1605617.98 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 4.0 +2023-02-06 23:02:13,154 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148906.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:02:35,643 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8310, 3.7579, 2.2646, 2.7069, 2.8418, 1.8818, 2.7080, 2.8031], + device='cuda:1'), covar=tensor([0.1939, 0.0343, 0.1123, 0.0833, 0.0774, 0.1478, 0.1147, 0.1202], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0239, 0.0330, 0.0306, 0.0301, 0.0334, 0.0345, 0.0320], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:02:38,068 INFO [train.py:901] (1/4) Epoch 19, batch 3450, loss[loss=0.2226, simple_loss=0.3057, pruned_loss=0.06972, over 8475.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.29, pruned_loss=0.0639, over 1604310.95 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 4.0 +2023-02-06 23:03:01,928 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148975.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:03:04,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.764e+02 2.340e+02 2.956e+02 3.727e+02 1.104e+03, threshold=5.912e+02, percent-clipped=3.0 +2023-02-06 23:03:14,135 INFO [train.py:901] (1/4) Epoch 19, batch 3500, loss[loss=0.199, simple_loss=0.281, pruned_loss=0.05854, over 8084.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2912, pruned_loss=0.0642, over 1609739.19 frames. ], batch size: 21, lr: 4.00e-03, grad_scale: 4.0 +2023-02-06 23:03:14,733 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 23:03:28,308 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=149013.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:03:33,360 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149020.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:03:35,952 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 23:03:38,850 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6642, 1.9107, 2.0888, 1.3486, 2.1754, 1.5102, 0.6499, 1.9009], + device='cuda:1'), covar=tensor([0.0605, 0.0362, 0.0282, 0.0566, 0.0384, 0.0787, 0.0807, 0.0297], + device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0379, 0.0328, 0.0436, 0.0361, 0.0526, 0.0379, 0.0406], + device='cuda:1'), out_proj_covar=tensor([1.1895e-04, 9.9959e-05, 8.6799e-05, 1.1606e-04, 9.5725e-05, 1.5048e-04, + 1.0273e-04, 1.0838e-04], device='cuda:1') +2023-02-06 23:03:48,890 INFO [train.py:901] (1/4) Epoch 19, batch 3550, loss[loss=0.261, simple_loss=0.3297, pruned_loss=0.09613, over 8239.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2901, pruned_loss=0.06371, over 1607998.81 frames. ], batch size: 24, lr: 4.00e-03, grad_scale: 4.0 +2023-02-06 23:03:50,365 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:03:51,746 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5406, 1.9701, 3.2032, 1.3264, 2.3182, 2.0623, 1.6445, 2.3272], + device='cuda:1'), covar=tensor([0.1938, 0.2571, 0.0896, 0.4593, 0.1879, 0.3014, 0.2326, 0.2319], + device='cuda:1'), in_proj_covar=tensor([0.0514, 0.0580, 0.0550, 0.0628, 0.0635, 0.0586, 0.0520, 0.0626], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:04:13,072 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=149078.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:04:13,638 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.461e+02 3.087e+02 3.824e+02 7.251e+02, threshold=6.175e+02, percent-clipped=6.0 +2023-02-06 23:04:25,654 INFO [train.py:901] (1/4) Epoch 19, batch 3600, loss[loss=0.1958, simple_loss=0.2784, pruned_loss=0.05657, over 8232.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2897, pruned_loss=0.06356, over 1602979.36 frames. ], batch size: 22, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:04:49,826 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:04:59,730 INFO [train.py:901] (1/4) Epoch 19, batch 3650, loss[loss=0.212, simple_loss=0.2822, pruned_loss=0.07085, over 7698.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2903, pruned_loss=0.06373, over 1604772.39 frames. ], batch size: 18, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:05:13,230 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149162.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:05:24,391 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.296e+02 2.731e+02 3.488e+02 6.725e+02, threshold=5.462e+02, percent-clipped=1.0 +2023-02-06 23:05:30,751 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149187.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:05:35,239 INFO [train.py:901] (1/4) Epoch 19, batch 3700, loss[loss=0.2132, simple_loss=0.2886, pruned_loss=0.06888, over 7541.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2909, pruned_loss=0.06437, over 1604323.92 frames. ], batch size: 18, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:05:35,425 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149193.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:05:38,923 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 23:06:02,794 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149231.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:06:10,472 INFO [train.py:901] (1/4) Epoch 19, batch 3750, loss[loss=0.2411, simple_loss=0.3159, pruned_loss=0.08316, over 8706.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2921, pruned_loss=0.06518, over 1609178.48 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:06:16,636 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3615, 1.4894, 4.5934, 1.8190, 4.0575, 3.8019, 4.1030, 3.9771], + device='cuda:1'), covar=tensor([0.0578, 0.4657, 0.0519, 0.3976, 0.1169, 0.0998, 0.0579, 0.0697], + device='cuda:1'), in_proj_covar=tensor([0.0598, 0.0633, 0.0672, 0.0606, 0.0684, 0.0593, 0.0584, 0.0648], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 23:06:19,380 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149256.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:06:34,560 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 2.530e+02 3.028e+02 3.831e+02 7.632e+02, threshold=6.056e+02, percent-clipped=6.0 +2023-02-06 23:06:44,214 INFO [train.py:901] (1/4) Epoch 19, batch 3800, loss[loss=0.2201, simple_loss=0.3044, pruned_loss=0.06788, over 8460.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2919, pruned_loss=0.06482, over 1613737.52 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:07:20,715 INFO [train.py:901] (1/4) Epoch 19, batch 3850, loss[loss=0.1789, simple_loss=0.2642, pruned_loss=0.04682, over 8108.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2922, pruned_loss=0.06506, over 1614760.63 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:07:25,676 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0315, 2.2705, 1.8834, 2.8787, 1.5675, 1.6042, 2.1391, 2.3031], + device='cuda:1'), covar=tensor([0.0771, 0.0774, 0.0931, 0.0372, 0.1027, 0.1316, 0.0869, 0.0772], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0196, 0.0248, 0.0212, 0.0205, 0.0245, 0.0251, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 23:07:42,390 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 23:07:45,095 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 2.409e+02 2.948e+02 3.728e+02 6.848e+02, threshold=5.896e+02, percent-clipped=3.0 +2023-02-06 23:07:48,726 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149384.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:07:54,628 INFO [train.py:901] (1/4) Epoch 19, batch 3900, loss[loss=0.1729, simple_loss=0.2525, pruned_loss=0.04671, over 7539.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2905, pruned_loss=0.06421, over 1614893.90 frames. ], batch size: 18, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:08:06,569 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149409.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:08:31,934 INFO [train.py:901] (1/4) Epoch 19, batch 3950, loss[loss=0.1771, simple_loss=0.2529, pruned_loss=0.05069, over 7700.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2907, pruned_loss=0.06417, over 1614325.45 frames. ], batch size: 18, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:08:36,308 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149449.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:08:39,079 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9766, 2.1688, 1.8050, 2.7527, 1.4584, 1.6127, 1.9096, 2.2052], + device='cuda:1'), covar=tensor([0.0725, 0.0739, 0.1014, 0.0422, 0.1042, 0.1312, 0.0920, 0.0755], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0198, 0.0250, 0.0214, 0.0206, 0.0247, 0.0254, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 23:08:53,034 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149474.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:08:56,239 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.497e+02 2.881e+02 4.050e+02 6.266e+02, threshold=5.763e+02, percent-clipped=1.0 +2023-02-06 23:09:05,731 INFO [train.py:901] (1/4) Epoch 19, batch 4000, loss[loss=0.2223, simple_loss=0.3001, pruned_loss=0.07221, over 8629.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.291, pruned_loss=0.06423, over 1618363.53 frames. ], batch size: 50, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:09:32,423 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:09:40,133 INFO [train.py:901] (1/4) Epoch 19, batch 4050, loss[loss=0.2109, simple_loss=0.3021, pruned_loss=0.0599, over 8525.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2908, pruned_loss=0.06439, over 1612853.82 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:09:41,068 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4724, 1.7362, 1.8964, 1.2151, 1.9181, 1.3806, 0.4239, 1.7100], + device='cuda:1'), covar=tensor([0.0496, 0.0370, 0.0311, 0.0495, 0.0413, 0.0841, 0.0754, 0.0252], + device='cuda:1'), in_proj_covar=tensor([0.0442, 0.0382, 0.0330, 0.0439, 0.0364, 0.0530, 0.0384, 0.0408], + device='cuda:1'), out_proj_covar=tensor([1.1971e-04, 1.0084e-04, 8.7247e-05, 1.1675e-04, 9.6446e-05, 1.5159e-04, + 1.0401e-04, 1.0912e-04], device='cuda:1') +2023-02-06 23:10:05,803 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.470e+02 3.003e+02 4.246e+02 8.728e+02, threshold=6.007e+02, percent-clipped=8.0 +2023-02-06 23:10:08,585 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:10:15,178 INFO [train.py:901] (1/4) Epoch 19, batch 4100, loss[loss=0.202, simple_loss=0.2954, pruned_loss=0.05434, over 8096.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2917, pruned_loss=0.06484, over 1613931.44 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:10:49,882 INFO [train.py:901] (1/4) Epoch 19, batch 4150, loss[loss=0.1989, simple_loss=0.2842, pruned_loss=0.05681, over 7966.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2922, pruned_loss=0.06528, over 1613112.44 frames. ], batch size: 21, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:10:51,538 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5100, 2.2203, 3.1888, 2.3912, 3.0095, 2.4102, 2.2938, 1.7344], + device='cuda:1'), covar=tensor([0.5005, 0.5115, 0.1941, 0.3767, 0.2477, 0.2970, 0.1734, 0.5609], + device='cuda:1'), in_proj_covar=tensor([0.0924, 0.0953, 0.0788, 0.0917, 0.0979, 0.0870, 0.0726, 0.0810], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 23:11:00,706 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.98 vs. limit=5.0 +2023-02-06 23:11:16,655 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.273e+02 2.791e+02 3.594e+02 5.057e+02, threshold=5.582e+02, percent-clipped=0.0 +2023-02-06 23:11:26,110 INFO [train.py:901] (1/4) Epoch 19, batch 4200, loss[loss=0.211, simple_loss=0.2849, pruned_loss=0.06853, over 7966.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2922, pruned_loss=0.06536, over 1611256.59 frames. ], batch size: 21, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:11:36,595 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 23:11:59,504 INFO [train.py:901] (1/4) Epoch 19, batch 4250, loss[loss=0.2045, simple_loss=0.2838, pruned_loss=0.06263, over 8081.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.293, pruned_loss=0.06602, over 1614768.31 frames. ], batch size: 21, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:12:00,930 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 23:12:14,481 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149764.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 23:12:25,320 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.327e+02 2.444e+02 3.025e+02 3.928e+02 1.033e+03, threshold=6.050e+02, percent-clipped=5.0 +2023-02-06 23:12:35,591 INFO [train.py:901] (1/4) Epoch 19, batch 4300, loss[loss=0.2252, simple_loss=0.2932, pruned_loss=0.07856, over 6014.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2917, pruned_loss=0.06498, over 1612529.88 frames. ], batch size: 13, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:13:08,872 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9916, 2.0333, 1.7797, 2.2413, 1.7054, 1.7741, 1.9158, 2.0998], + device='cuda:1'), covar=tensor([0.0622, 0.0663, 0.0788, 0.0602, 0.0813, 0.0969, 0.0657, 0.0638], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0199, 0.0248, 0.0213, 0.0206, 0.0247, 0.0253, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 23:13:10,096 INFO [train.py:901] (1/4) Epoch 19, batch 4350, loss[loss=0.1754, simple_loss=0.2485, pruned_loss=0.05122, over 7447.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2907, pruned_loss=0.06451, over 1612666.40 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:13:33,182 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 23:13:33,236 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=149876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:13:33,384 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.7567, 0.8504, 0.8084, 0.4588, 0.8279, 0.6685, 0.0789, 0.8322], + device='cuda:1'), covar=tensor([0.0281, 0.0264, 0.0220, 0.0401, 0.0254, 0.0614, 0.0566, 0.0220], + device='cuda:1'), in_proj_covar=tensor([0.0442, 0.0383, 0.0330, 0.0438, 0.0365, 0.0530, 0.0384, 0.0408], + device='cuda:1'), out_proj_covar=tensor([1.1955e-04, 1.0115e-04, 8.7273e-05, 1.1623e-04, 9.6675e-05, 1.5135e-04, + 1.0400e-04, 1.0905e-04], device='cuda:1') +2023-02-06 23:13:35,197 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.529e+02 2.416e+02 2.972e+02 3.761e+02 1.184e+03, threshold=5.944e+02, percent-clipped=4.0 +2023-02-06 23:13:44,577 INFO [train.py:901] (1/4) Epoch 19, batch 4400, loss[loss=0.2301, simple_loss=0.3073, pruned_loss=0.0764, over 8533.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2905, pruned_loss=0.06449, over 1607226.14 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:14:09,979 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=149927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:14:14,583 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 23:14:20,892 INFO [train.py:901] (1/4) Epoch 19, batch 4450, loss[loss=0.2393, simple_loss=0.3121, pruned_loss=0.08327, over 8674.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2913, pruned_loss=0.0653, over 1605435.51 frames. ], batch size: 31, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:14:44,886 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.522e+02 2.925e+02 4.193e+02 1.036e+03, threshold=5.849e+02, percent-clipped=7.0 +2023-02-06 23:14:53,307 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149991.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:14:54,478 INFO [train.py:901] (1/4) Epoch 19, batch 4500, loss[loss=0.1847, simple_loss=0.2642, pruned_loss=0.05259, over 7816.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2903, pruned_loss=0.06432, over 1609230.14 frames. ], batch size: 20, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:15:08,410 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 23:15:11,309 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9186, 1.3400, 3.0851, 1.3121, 2.0504, 3.3225, 3.5126, 2.7932], + device='cuda:1'), covar=tensor([0.1138, 0.1888, 0.0386, 0.2353, 0.1144, 0.0273, 0.0469, 0.0633], + device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0317, 0.0287, 0.0313, 0.0302, 0.0265, 0.0407, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:15:31,633 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:15:32,129 INFO [train.py:901] (1/4) Epoch 19, batch 4550, loss[loss=0.2153, simple_loss=0.2822, pruned_loss=0.07419, over 7238.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.291, pruned_loss=0.06485, over 1612617.97 frames. ], batch size: 16, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:15:56,361 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.399e+02 2.811e+02 3.428e+02 5.502e+02, threshold=5.622e+02, percent-clipped=0.0 +2023-02-06 23:16:05,766 INFO [train.py:901] (1/4) Epoch 19, batch 4600, loss[loss=0.1902, simple_loss=0.2573, pruned_loss=0.06151, over 7542.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2903, pruned_loss=0.06431, over 1611309.22 frames. ], batch size: 18, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:16:08,470 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150097.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:16:10,591 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9999, 2.1286, 1.8665, 2.7656, 1.3299, 1.6822, 1.9378, 2.1437], + device='cuda:1'), covar=tensor([0.0777, 0.0825, 0.0963, 0.0378, 0.1078, 0.1258, 0.0896, 0.0772], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0198, 0.0248, 0.0213, 0.0206, 0.0248, 0.0254, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 23:16:12,113 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.26 vs. limit=5.0 +2023-02-06 23:16:15,961 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=150108.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 23:16:40,343 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5868, 1.8441, 3.0739, 1.4231, 2.2470, 1.9888, 1.6181, 2.1645], + device='cuda:1'), covar=tensor([0.1864, 0.2675, 0.0748, 0.4564, 0.1843, 0.3036, 0.2293, 0.2257], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0587, 0.0557, 0.0634, 0.0643, 0.0592, 0.0528, 0.0635], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:16:41,502 INFO [train.py:901] (1/4) Epoch 19, batch 4650, loss[loss=0.2303, simple_loss=0.3139, pruned_loss=0.0733, over 8501.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2898, pruned_loss=0.06424, over 1610001.25 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:17:06,564 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.474e+02 2.856e+02 3.464e+02 8.049e+02, threshold=5.712e+02, percent-clipped=3.0 +2023-02-06 23:17:16,089 INFO [train.py:901] (1/4) Epoch 19, batch 4700, loss[loss=0.2388, simple_loss=0.3255, pruned_loss=0.07607, over 8333.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2893, pruned_loss=0.06412, over 1609757.85 frames. ], batch size: 26, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:17:21,222 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0983, 2.1690, 1.9566, 2.8647, 1.4110, 1.6968, 1.9530, 2.2711], + device='cuda:1'), covar=tensor([0.0717, 0.0780, 0.0854, 0.0352, 0.1072, 0.1268, 0.0949, 0.0740], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0197, 0.0246, 0.0212, 0.0205, 0.0247, 0.0253, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 23:17:36,661 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150223.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 23:17:49,661 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3814, 4.3315, 3.9000, 1.9365, 3.8313, 4.0971, 3.9601, 3.7262], + device='cuda:1'), covar=tensor([0.0847, 0.0597, 0.1277, 0.5231, 0.0978, 0.1057, 0.1363, 0.0932], + device='cuda:1'), in_proj_covar=tensor([0.0517, 0.0428, 0.0426, 0.0528, 0.0416, 0.0433, 0.0411, 0.0375], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:17:50,825 INFO [train.py:901] (1/4) Epoch 19, batch 4750, loss[loss=0.1853, simple_loss=0.2647, pruned_loss=0.05297, over 7791.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2892, pruned_loss=0.06389, over 1608571.86 frames. ], batch size: 19, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:17:53,798 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:18:12,320 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150272.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:18:13,465 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 23:18:15,513 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 23:18:16,855 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.315e+02 2.829e+02 3.523e+02 6.730e+02, threshold=5.657e+02, percent-clipped=3.0 +2023-02-06 23:18:25,810 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150292.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:18:26,369 INFO [train.py:901] (1/4) Epoch 19, batch 4800, loss[loss=0.1847, simple_loss=0.28, pruned_loss=0.04471, over 8423.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2888, pruned_loss=0.06322, over 1610217.59 frames. ], batch size: 27, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:18:29,938 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150298.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:18:46,562 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150323.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:19:00,012 INFO [train.py:901] (1/4) Epoch 19, batch 4850, loss[loss=0.2142, simple_loss=0.2976, pruned_loss=0.06545, over 8562.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2884, pruned_loss=0.06283, over 1614056.42 frames. ], batch size: 31, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:19:05,355 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 23:19:27,018 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.476e+02 2.899e+02 3.621e+02 6.951e+02, threshold=5.799e+02, percent-clipped=6.0 +2023-02-06 23:19:36,191 INFO [train.py:901] (1/4) Epoch 19, batch 4900, loss[loss=0.2423, simple_loss=0.3208, pruned_loss=0.08184, over 8429.00 frames. ], tot_loss[loss=0.208, simple_loss=0.289, pruned_loss=0.06347, over 1615662.63 frames. ], batch size: 29, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:19:39,302 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-02-06 23:20:07,720 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=150441.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:20:08,504 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0640, 2.3931, 1.8385, 2.8795, 1.3957, 1.6648, 2.0102, 2.3844], + device='cuda:1'), covar=tensor([0.0725, 0.0678, 0.0939, 0.0368, 0.1120, 0.1318, 0.0927, 0.0774], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0198, 0.0248, 0.0213, 0.0205, 0.0248, 0.0254, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-06 23:20:09,004 INFO [train.py:901] (1/4) Epoch 19, batch 4950, loss[loss=0.2071, simple_loss=0.2867, pruned_loss=0.06377, over 8479.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2903, pruned_loss=0.0645, over 1616872.40 frames. ], batch size: 29, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:20:33,699 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.356e+02 2.775e+02 3.573e+02 1.033e+03, threshold=5.550e+02, percent-clipped=4.0 +2023-02-06 23:20:33,939 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150479.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 23:20:41,454 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5232, 1.7903, 4.4050, 2.0654, 2.4446, 4.9572, 5.0501, 4.2844], + device='cuda:1'), covar=tensor([0.1067, 0.1646, 0.0241, 0.1901, 0.1191, 0.0180, 0.0479, 0.0538], + device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0319, 0.0289, 0.0313, 0.0304, 0.0267, 0.0408, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:20:43,985 INFO [train.py:901] (1/4) Epoch 19, batch 5000, loss[loss=0.2392, simple_loss=0.3275, pruned_loss=0.07543, over 8438.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2906, pruned_loss=0.06426, over 1619376.15 frames. ], batch size: 29, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:20:50,110 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6143, 2.5570, 1.9301, 2.2056, 2.0541, 1.6592, 2.0137, 2.1756], + device='cuda:1'), covar=tensor([0.1564, 0.0389, 0.1122, 0.0633, 0.0769, 0.1478, 0.1062, 0.1037], + device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0237, 0.0326, 0.0305, 0.0301, 0.0331, 0.0341, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:20:52,071 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150504.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 23:21:17,801 INFO [train.py:901] (1/4) Epoch 19, batch 5050, loss[loss=0.219, simple_loss=0.2945, pruned_loss=0.07172, over 8609.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.291, pruned_loss=0.06491, over 1619262.64 frames. ], batch size: 31, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:21:25,912 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150555.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:21:26,618 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150556.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:21:40,937 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 23:21:41,604 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.501e+02 3.000e+02 3.972e+02 7.212e+02, threshold=5.999e+02, percent-clipped=3.0 +2023-02-06 23:21:51,786 INFO [train.py:901] (1/4) Epoch 19, batch 5100, loss[loss=0.2986, simple_loss=0.353, pruned_loss=0.1221, over 6835.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2914, pruned_loss=0.06531, over 1617027.67 frames. ], batch size: 72, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:22:23,318 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=150636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:22:27,868 INFO [train.py:901] (1/4) Epoch 19, batch 5150, loss[loss=0.1995, simple_loss=0.2892, pruned_loss=0.05491, over 7810.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2903, pruned_loss=0.06477, over 1612344.28 frames. ], batch size: 20, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:22:51,867 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.708e+02 2.510e+02 3.215e+02 4.688e+02 9.098e+02, threshold=6.429e+02, percent-clipped=11.0 +2023-02-06 23:23:01,331 INFO [train.py:901] (1/4) Epoch 19, batch 5200, loss[loss=0.2243, simple_loss=0.3161, pruned_loss=0.06625, over 8593.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2895, pruned_loss=0.06394, over 1614911.86 frames. ], batch size: 31, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:23:28,756 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6159, 1.6402, 2.2610, 1.4697, 1.1751, 2.3055, 0.4399, 1.3859], + device='cuda:1'), covar=tensor([0.2295, 0.1463, 0.0391, 0.1721, 0.3004, 0.0387, 0.2254, 0.1619], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0193, 0.0123, 0.0220, 0.0268, 0.0132, 0.0169, 0.0187], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 23:23:38,104 INFO [train.py:901] (1/4) Epoch 19, batch 5250, loss[loss=0.2014, simple_loss=0.2898, pruned_loss=0.05653, over 8472.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2898, pruned_loss=0.06395, over 1613362.81 frames. ], batch size: 29, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:23:40,271 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7653, 1.8238, 2.2508, 1.6521, 1.3067, 2.2792, 0.5225, 1.4377], + device='cuda:1'), covar=tensor([0.1464, 0.1039, 0.0386, 0.1114, 0.2457, 0.0353, 0.1896, 0.1186], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0193, 0.0123, 0.0220, 0.0267, 0.0132, 0.0169, 0.0187], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 23:23:40,650 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 23:23:42,656 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:23:43,374 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150751.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:24:01,540 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.351e+02 2.565e+02 3.080e+02 4.191e+02 1.354e+03, threshold=6.160e+02, percent-clipped=9.0 +2023-02-06 23:24:10,893 INFO [train.py:901] (1/4) Epoch 19, batch 5300, loss[loss=0.243, simple_loss=0.3257, pruned_loss=0.08013, over 8741.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2908, pruned_loss=0.0651, over 1611586.34 frames. ], batch size: 30, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:24:23,655 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150812.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:24:41,664 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150837.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:24:46,253 INFO [train.py:901] (1/4) Epoch 19, batch 5350, loss[loss=0.2287, simple_loss=0.2961, pruned_loss=0.0807, over 8076.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2907, pruned_loss=0.06506, over 1605443.64 frames. ], batch size: 21, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:24:58,168 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 23:25:10,967 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.499e+02 2.979e+02 3.723e+02 8.863e+02, threshold=5.959e+02, percent-clipped=1.0 +2023-02-06 23:25:20,520 INFO [train.py:901] (1/4) Epoch 19, batch 5400, loss[loss=0.216, simple_loss=0.3039, pruned_loss=0.06401, over 8328.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2925, pruned_loss=0.06582, over 1610196.43 frames. ], batch size: 25, lr: 3.98e-03, grad_scale: 16.0 +2023-02-06 23:25:24,750 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=150899.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:25:27,896 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 23:25:37,927 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150918.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:25:55,457 INFO [train.py:901] (1/4) Epoch 19, batch 5450, loss[loss=0.2453, simple_loss=0.3231, pruned_loss=0.08371, over 8691.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2919, pruned_loss=0.06546, over 1611779.95 frames. ], batch size: 34, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:26:22,623 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.347e+02 2.658e+02 3.430e+02 7.604e+02, threshold=5.316e+02, percent-clipped=2.0 +2023-02-06 23:26:26,887 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 +2023-02-06 23:26:28,462 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 23:26:31,932 INFO [train.py:901] (1/4) Epoch 19, batch 5500, loss[loss=0.2103, simple_loss=0.2917, pruned_loss=0.06448, over 8143.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.291, pruned_loss=0.06504, over 1611778.56 frames. ], batch size: 22, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:26:41,796 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151007.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:26:46,662 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151014.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:26:48,801 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.41 vs. limit=5.0 +2023-02-06 23:26:58,985 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:27:06,223 INFO [train.py:901] (1/4) Epoch 19, batch 5550, loss[loss=0.2215, simple_loss=0.2879, pruned_loss=0.07761, over 7652.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2897, pruned_loss=0.06418, over 1611996.79 frames. ], batch size: 19, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:27:27,974 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7053, 1.6686, 2.3462, 1.5546, 1.2908, 2.3109, 0.4971, 1.3927], + device='cuda:1'), covar=tensor([0.1967, 0.1356, 0.0404, 0.1573, 0.2868, 0.0430, 0.2405, 0.1628], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0194, 0.0124, 0.0222, 0.0269, 0.0133, 0.0169, 0.0188], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 23:27:32,496 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.454e+02 3.027e+02 4.195e+02 6.901e+02, threshold=6.054e+02, percent-clipped=7.0 +2023-02-06 23:27:42,411 INFO [train.py:901] (1/4) Epoch 19, batch 5600, loss[loss=0.1887, simple_loss=0.2786, pruned_loss=0.0494, over 8093.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2903, pruned_loss=0.06457, over 1608554.17 frames. ], batch size: 21, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:27:43,136 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=151094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:27:47,848 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7112, 1.7109, 1.9993, 1.6052, 1.2155, 1.7954, 2.3288, 2.1415], + device='cuda:1'), covar=tensor([0.0515, 0.1560, 0.2021, 0.1797, 0.0727, 0.1829, 0.0692, 0.0645], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0158, 0.0099, 0.0161, 0.0112, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 23:28:15,312 INFO [train.py:901] (1/4) Epoch 19, batch 5650, loss[loss=0.1986, simple_loss=0.28, pruned_loss=0.05858, over 7239.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2912, pruned_loss=0.06485, over 1607929.66 frames. ], batch size: 16, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:28:25,140 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4356, 2.1664, 2.8821, 2.3725, 2.8048, 2.4193, 2.1761, 1.7763], + device='cuda:1'), covar=tensor([0.4308, 0.4233, 0.1607, 0.2975, 0.1984, 0.2471, 0.1677, 0.4324], + device='cuda:1'), in_proj_covar=tensor([0.0934, 0.0960, 0.0789, 0.0924, 0.0984, 0.0874, 0.0740, 0.0815], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 23:28:31,571 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 23:28:39,668 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.685e+02 3.149e+02 3.866e+02 8.044e+02, threshold=6.298e+02, percent-clipped=3.0 +2023-02-06 23:28:50,395 INFO [train.py:901] (1/4) Epoch 19, batch 5700, loss[loss=0.1775, simple_loss=0.2584, pruned_loss=0.04827, over 8081.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2899, pruned_loss=0.06463, over 1605890.73 frames. ], batch size: 21, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:28:57,633 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151202.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:29:02,316 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151209.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:29:11,910 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-02-06 23:29:24,756 INFO [train.py:901] (1/4) Epoch 19, batch 5750, loss[loss=0.2133, simple_loss=0.3021, pruned_loss=0.06223, over 8565.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2888, pruned_loss=0.06433, over 1603144.27 frames. ], batch size: 39, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:29:36,097 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 23:29:37,494 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=151262.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:29:41,748 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6902, 2.2933, 2.9557, 2.4827, 2.9245, 2.5225, 2.4202, 2.1691], + device='cuda:1'), covar=tensor([0.3296, 0.3779, 0.1432, 0.2683, 0.1751, 0.2359, 0.1418, 0.3594], + device='cuda:1'), in_proj_covar=tensor([0.0929, 0.0956, 0.0786, 0.0919, 0.0979, 0.0870, 0.0735, 0.0811], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 23:29:43,023 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151270.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:29:48,610 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.389e+02 2.918e+02 3.727e+02 7.769e+02, threshold=5.836e+02, percent-clipped=3.0 +2023-02-06 23:29:58,863 INFO [train.py:901] (1/4) Epoch 19, batch 5800, loss[loss=0.24, simple_loss=0.3239, pruned_loss=0.0781, over 8480.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2896, pruned_loss=0.06473, over 1599155.64 frames. ], batch size: 25, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:30:00,355 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151295.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:30:04,359 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151300.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:30:05,086 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5050, 2.0278, 3.0763, 1.3367, 2.3341, 1.8561, 1.6618, 2.1456], + device='cuda:1'), covar=tensor([0.2139, 0.2417, 0.0879, 0.4725, 0.1953, 0.3422, 0.2441, 0.2509], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0580, 0.0552, 0.0629, 0.0637, 0.0587, 0.0521, 0.0626], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:30:16,600 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151316.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:30:34,589 INFO [train.py:901] (1/4) Epoch 19, batch 5850, loss[loss=0.2067, simple_loss=0.2948, pruned_loss=0.05932, over 8256.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2908, pruned_loss=0.06496, over 1604292.96 frames. ], batch size: 24, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:30:57,562 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151377.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:30:58,661 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.316e+02 2.176e+02 2.714e+02 3.221e+02 1.387e+03, threshold=5.429e+02, percent-clipped=3.0 +2023-02-06 23:31:08,074 INFO [train.py:901] (1/4) Epoch 19, batch 5900, loss[loss=0.1927, simple_loss=0.2718, pruned_loss=0.05682, over 7799.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2908, pruned_loss=0.06475, over 1607426.44 frames. ], batch size: 20, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:31:15,414 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 23:31:44,452 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5936, 1.9819, 3.3188, 1.4009, 2.4268, 1.9704, 1.6982, 2.4004], + device='cuda:1'), covar=tensor([0.1824, 0.2386, 0.0807, 0.4361, 0.1756, 0.3092, 0.2145, 0.2288], + device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0583, 0.0555, 0.0632, 0.0640, 0.0590, 0.0523, 0.0629], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:31:44,896 INFO [train.py:901] (1/4) Epoch 19, batch 5950, loss[loss=0.2601, simple_loss=0.3351, pruned_loss=0.09255, over 8355.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2907, pruned_loss=0.06449, over 1607090.72 frames. ], batch size: 24, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:31:59,862 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151465.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:32:09,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.424e+02 3.104e+02 3.851e+02 8.156e+02, threshold=6.208e+02, percent-clipped=3.0 +2023-02-06 23:32:16,803 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:32:18,556 INFO [train.py:901] (1/4) Epoch 19, batch 6000, loss[loss=0.1878, simple_loss=0.2605, pruned_loss=0.05754, over 7731.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2913, pruned_loss=0.06495, over 1612389.56 frames. ], batch size: 18, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:32:18,557 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 23:32:32,009 INFO [train.py:935] (1/4) Epoch 19, validation: loss=0.1763, simple_loss=0.2764, pruned_loss=0.03805, over 944034.00 frames. +2023-02-06 23:32:32,012 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 23:32:56,730 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6454, 1.9292, 2.1008, 1.2775, 2.1550, 1.5180, 0.7175, 1.8311], + device='cuda:1'), covar=tensor([0.0684, 0.0407, 0.0340, 0.0683, 0.0450, 0.0951, 0.0971, 0.0358], + device='cuda:1'), in_proj_covar=tensor([0.0444, 0.0380, 0.0332, 0.0439, 0.0368, 0.0527, 0.0386, 0.0406], + device='cuda:1'), out_proj_covar=tensor([1.2020e-04, 1.0041e-04, 8.7768e-05, 1.1665e-04, 9.7573e-05, 1.5037e-04, + 1.0445e-04, 1.0841e-04], device='cuda:1') +2023-02-06 23:32:59,489 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9173, 2.3122, 3.6255, 2.0601, 1.8054, 3.4662, 0.8088, 2.1126], + device='cuda:1'), covar=tensor([0.1271, 0.1341, 0.0207, 0.1718, 0.2787, 0.0321, 0.2297, 0.1455], + device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0193, 0.0123, 0.0221, 0.0268, 0.0133, 0.0168, 0.0187], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 23:33:06,951 INFO [train.py:901] (1/4) Epoch 19, batch 6050, loss[loss=0.2041, simple_loss=0.2894, pruned_loss=0.05945, over 8146.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2926, pruned_loss=0.06597, over 1607989.76 frames. ], batch size: 22, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:33:09,101 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=151546.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:33:32,578 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.535e+02 3.172e+02 3.888e+02 8.825e+02, threshold=6.343e+02, percent-clipped=4.0 +2023-02-06 23:33:42,769 INFO [train.py:901] (1/4) Epoch 19, batch 6100, loss[loss=0.1876, simple_loss=0.2767, pruned_loss=0.04928, over 8250.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2924, pruned_loss=0.06541, over 1612246.32 frames. ], batch size: 22, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:33:56,036 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2005, 1.4605, 3.4917, 1.4071, 2.2932, 3.8262, 3.8957, 3.3000], + device='cuda:1'), covar=tensor([0.0940, 0.1693, 0.0286, 0.2019, 0.1020, 0.0194, 0.0418, 0.0506], + device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0319, 0.0286, 0.0311, 0.0303, 0.0262, 0.0405, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:34:07,689 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 23:34:10,860 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151633.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:17,587 INFO [train.py:901] (1/4) Epoch 19, batch 6150, loss[loss=0.1986, simple_loss=0.2875, pruned_loss=0.05484, over 8227.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2909, pruned_loss=0.06481, over 1611554.32 frames. ], batch size: 22, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:34:18,368 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=151644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:28,851 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151658.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:30,136 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=151660.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:30,975 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:43,585 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.320e+02 2.846e+02 3.654e+02 5.745e+02, threshold=5.693e+02, percent-clipped=0.0 +2023-02-06 23:34:51,571 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-06 23:34:53,948 INFO [train.py:901] (1/4) Epoch 19, batch 6200, loss[loss=0.2271, simple_loss=0.3029, pruned_loss=0.07566, over 8327.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2905, pruned_loss=0.06464, over 1612498.80 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:35:02,701 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151706.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:35:28,542 INFO [train.py:901] (1/4) Epoch 19, batch 6250, loss[loss=0.2243, simple_loss=0.2996, pruned_loss=0.07452, over 8623.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2909, pruned_loss=0.06501, over 1611676.39 frames. ], batch size: 39, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:35:31,304 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8474, 1.6348, 5.8921, 2.2320, 5.3707, 5.0007, 5.4511, 5.3251], + device='cuda:1'), covar=tensor([0.0314, 0.4575, 0.0383, 0.3538, 0.0798, 0.0847, 0.0441, 0.0426], + device='cuda:1'), in_proj_covar=tensor([0.0593, 0.0627, 0.0669, 0.0602, 0.0678, 0.0583, 0.0584, 0.0643], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:35:39,356 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151759.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:35:50,915 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151775.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:35:53,502 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.555e+02 3.246e+02 4.070e+02 8.549e+02, threshold=6.492e+02, percent-clipped=6.0 +2023-02-06 23:36:03,713 INFO [train.py:901] (1/4) Epoch 19, batch 6300, loss[loss=0.2291, simple_loss=0.3104, pruned_loss=0.07391, over 8360.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2925, pruned_loss=0.06548, over 1615424.25 frames. ], batch size: 24, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:36:22,223 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151819.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:36:27,269 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 23:36:39,103 INFO [train.py:901] (1/4) Epoch 19, batch 6350, loss[loss=0.2141, simple_loss=0.2888, pruned_loss=0.06973, over 7972.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2927, pruned_loss=0.06589, over 1615387.30 frames. ], batch size: 21, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:37:03,145 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.433e+02 2.376e+02 2.921e+02 3.593e+02 6.855e+02, threshold=5.841e+02, percent-clipped=1.0 +2023-02-06 23:37:10,746 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1219, 1.8846, 3.5231, 1.5836, 2.3065, 3.8841, 3.9801, 3.3612], + device='cuda:1'), covar=tensor([0.1032, 0.1450, 0.0315, 0.2035, 0.1128, 0.0230, 0.0623, 0.0553], + device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0316, 0.0284, 0.0310, 0.0301, 0.0261, 0.0405, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:37:13,211 INFO [train.py:901] (1/4) Epoch 19, batch 6400, loss[loss=0.185, simple_loss=0.2768, pruned_loss=0.0466, over 8201.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2929, pruned_loss=0.06585, over 1615760.27 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:37:30,643 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151917.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:37:48,047 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151942.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:37:48,526 INFO [train.py:901] (1/4) Epoch 19, batch 6450, loss[loss=0.203, simple_loss=0.283, pruned_loss=0.0615, over 8082.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2924, pruned_loss=0.06575, over 1615385.87 frames. ], batch size: 21, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:38:13,510 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.409e+02 2.943e+02 3.710e+02 6.232e+02, threshold=5.887e+02, percent-clipped=1.0 +2023-02-06 23:38:23,089 INFO [train.py:901] (1/4) Epoch 19, batch 6500, loss[loss=0.1435, simple_loss=0.2287, pruned_loss=0.02917, over 5536.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2916, pruned_loss=0.06552, over 1608758.95 frames. ], batch size: 12, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:38:40,015 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152015.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:38:51,598 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152031.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:38:58,592 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152040.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:38:59,457 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 23:39:00,449 INFO [train.py:901] (1/4) Epoch 19, batch 6550, loss[loss=0.2065, simple_loss=0.2995, pruned_loss=0.05678, over 8254.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2918, pruned_loss=0.06538, over 1610089.38 frames. ], batch size: 24, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:39:04,937 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152050.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:39:09,227 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152056.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:39:21,623 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 23:39:24,907 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.379e+02 2.761e+02 3.695e+02 7.678e+02, threshold=5.522e+02, percent-clipped=3.0 +2023-02-06 23:39:34,313 INFO [train.py:901] (1/4) Epoch 19, batch 6600, loss[loss=0.1989, simple_loss=0.2718, pruned_loss=0.06301, over 7798.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2915, pruned_loss=0.06524, over 1606148.85 frames. ], batch size: 19, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:39:39,614 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 23:40:09,009 INFO [train.py:901] (1/4) Epoch 19, batch 6650, loss[loss=0.2228, simple_loss=0.3027, pruned_loss=0.07149, over 8038.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2931, pruned_loss=0.06611, over 1609876.97 frames. ], batch size: 22, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:40:23,465 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152163.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:40:24,953 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:40:34,183 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 2.686e+02 3.265e+02 3.895e+02 8.931e+02, threshold=6.531e+02, percent-clipped=7.0 +2023-02-06 23:40:44,526 INFO [train.py:901] (1/4) Epoch 19, batch 6700, loss[loss=0.2564, simple_loss=0.3326, pruned_loss=0.09006, over 8439.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2931, pruned_loss=0.06588, over 1614167.27 frames. ], batch size: 29, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:41:10,479 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9479, 1.5892, 3.4699, 1.5675, 2.4139, 3.8689, 3.8871, 3.3137], + device='cuda:1'), covar=tensor([0.1270, 0.1782, 0.0361, 0.2127, 0.1139, 0.0214, 0.0514, 0.0569], + device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0318, 0.0284, 0.0310, 0.0301, 0.0260, 0.0405, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:41:16,176 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6144, 1.6968, 1.7344, 1.4252, 1.8417, 1.4283, 0.8791, 1.6813], + device='cuda:1'), covar=tensor([0.0471, 0.0354, 0.0224, 0.0424, 0.0338, 0.0644, 0.0681, 0.0220], + device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0377, 0.0330, 0.0436, 0.0363, 0.0522, 0.0382, 0.0402], + device='cuda:1'), out_proj_covar=tensor([1.1824e-04, 9.9495e-05, 8.7267e-05, 1.1562e-04, 9.6266e-05, 1.4874e-04, + 1.0334e-04, 1.0728e-04], device='cuda:1') +2023-02-06 23:41:19,464 INFO [train.py:901] (1/4) Epoch 19, batch 6750, loss[loss=0.1734, simple_loss=0.2606, pruned_loss=0.04312, over 8124.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2926, pruned_loss=0.06541, over 1613370.74 frames. ], batch size: 22, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:41:19,611 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152243.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:41:38,483 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8369, 1.7632, 1.9341, 1.7712, 1.1405, 1.7992, 2.2863, 2.0514], + device='cuda:1'), covar=tensor([0.0448, 0.1198, 0.1573, 0.1319, 0.0638, 0.1346, 0.0609, 0.0588], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0157, 0.0099, 0.0160, 0.0112, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 23:41:44,685 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:41:45,124 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.378e+02 2.909e+02 3.491e+02 6.752e+02, threshold=5.817e+02, percent-clipped=2.0 +2023-02-06 23:41:53,502 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:41:54,060 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 23:41:54,743 INFO [train.py:901] (1/4) Epoch 19, batch 6800, loss[loss=0.1959, simple_loss=0.2787, pruned_loss=0.05653, over 8359.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2924, pruned_loss=0.06527, over 1614960.57 frames. ], batch size: 24, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:41:55,685 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6087, 1.8963, 3.0304, 1.4270, 2.2792, 2.0363, 1.6463, 2.3873], + device='cuda:1'), covar=tensor([0.1901, 0.2499, 0.0781, 0.4419, 0.1742, 0.3248, 0.2317, 0.1913], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0581, 0.0549, 0.0627, 0.0638, 0.0586, 0.0521, 0.0627], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:42:29,090 INFO [train.py:901] (1/4) Epoch 19, batch 6850, loss[loss=0.2109, simple_loss=0.2873, pruned_loss=0.06727, over 8035.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2919, pruned_loss=0.06485, over 1614513.87 frames. ], batch size: 22, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:42:43,979 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 23:42:54,742 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.344e+02 3.012e+02 3.839e+02 8.073e+02, threshold=6.025e+02, percent-clipped=5.0 +2023-02-06 23:42:56,514 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 23:43:00,375 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7998, 5.9471, 5.1293, 2.1691, 5.2278, 5.4652, 5.4581, 5.3340], + device='cuda:1'), covar=tensor([0.0486, 0.0381, 0.0933, 0.4752, 0.0694, 0.0875, 0.0993, 0.0517], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0428, 0.0431, 0.0535, 0.0418, 0.0433, 0.0414, 0.0376], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:43:02,848 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 23:43:05,118 INFO [train.py:901] (1/4) Epoch 19, batch 6900, loss[loss=0.2049, simple_loss=0.2899, pruned_loss=0.05993, over 8369.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2916, pruned_loss=0.0643, over 1615840.94 frames. ], batch size: 24, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:43:18,941 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 23:43:25,518 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152421.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:43:40,396 INFO [train.py:901] (1/4) Epoch 19, batch 6950, loss[loss=0.2518, simple_loss=0.3189, pruned_loss=0.09237, over 6771.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2923, pruned_loss=0.06487, over 1613713.79 frames. ], batch size: 73, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:43:42,624 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152446.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:43:43,491 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 +2023-02-06 23:43:53,794 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 23:43:53,935 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152463.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:44:05,256 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.443e+02 3.132e+02 3.706e+02 6.613e+02, threshold=6.264e+02, percent-clipped=2.0 +2023-02-06 23:44:14,628 INFO [train.py:901] (1/4) Epoch 19, batch 7000, loss[loss=0.2286, simple_loss=0.3067, pruned_loss=0.07521, over 6775.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2928, pruned_loss=0.06537, over 1614823.36 frames. ], batch size: 71, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:44:24,958 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7372, 1.4424, 4.9042, 1.7026, 4.4305, 4.0245, 4.4459, 4.3519], + device='cuda:1'), covar=tensor([0.0444, 0.4284, 0.0394, 0.3817, 0.0970, 0.0929, 0.0438, 0.0543], + device='cuda:1'), in_proj_covar=tensor([0.0593, 0.0626, 0.0667, 0.0603, 0.0682, 0.0585, 0.0583, 0.0645], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 23:44:44,335 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152534.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:44:49,244 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1990, 2.0133, 2.7365, 2.2300, 2.6447, 2.2245, 2.0256, 1.4319], + device='cuda:1'), covar=tensor([0.5569, 0.5000, 0.1908, 0.3725, 0.2537, 0.3181, 0.2129, 0.5584], + device='cuda:1'), in_proj_covar=tensor([0.0930, 0.0960, 0.0786, 0.0922, 0.0981, 0.0874, 0.0738, 0.0816], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-06 23:44:51,104 INFO [train.py:901] (1/4) Epoch 19, batch 7050, loss[loss=0.2121, simple_loss=0.3023, pruned_loss=0.06101, over 8329.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2919, pruned_loss=0.06416, over 1618802.78 frames. ], batch size: 26, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:44:59,579 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0768, 3.9420, 2.4967, 2.9506, 3.0669, 2.2403, 2.9912, 3.1694], + device='cuda:1'), covar=tensor([0.1629, 0.0298, 0.1163, 0.0661, 0.0694, 0.1382, 0.0991, 0.1066], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0235, 0.0330, 0.0305, 0.0303, 0.0334, 0.0344, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:45:01,133 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 +2023-02-06 23:45:02,279 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152559.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:45:15,726 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 2.420e+02 2.800e+02 3.429e+02 5.549e+02, threshold=5.599e+02, percent-clipped=0.0 +2023-02-06 23:45:21,280 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:45:25,399 INFO [train.py:901] (1/4) Epoch 19, batch 7100, loss[loss=0.1871, simple_loss=0.2559, pruned_loss=0.05914, over 7241.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.291, pruned_loss=0.06392, over 1617017.99 frames. ], batch size: 16, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:45:30,681 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152600.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:45:35,439 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:45:56,369 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:46:01,875 INFO [train.py:901] (1/4) Epoch 19, batch 7150, loss[loss=0.1597, simple_loss=0.2487, pruned_loss=0.03535, over 7815.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2926, pruned_loss=0.0648, over 1616059.42 frames. ], batch size: 20, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:46:26,711 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4197, 1.3735, 1.7489, 1.1939, 1.0990, 1.7159, 0.2070, 1.1203], + device='cuda:1'), covar=tensor([0.1966, 0.1549, 0.0402, 0.1067, 0.2747, 0.0509, 0.2278, 0.1406], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0194, 0.0123, 0.0220, 0.0266, 0.0132, 0.0168, 0.0186], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 23:46:27,177 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.485e+02 2.441e+02 2.885e+02 3.630e+02 1.043e+03, threshold=5.770e+02, percent-clipped=5.0 +2023-02-06 23:46:36,617 INFO [train.py:901] (1/4) Epoch 19, batch 7200, loss[loss=0.197, simple_loss=0.2854, pruned_loss=0.05429, over 8444.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2931, pruned_loss=0.06551, over 1618642.66 frames. ], batch size: 29, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:46:42,735 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152702.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:47:12,598 INFO [train.py:901] (1/4) Epoch 19, batch 7250, loss[loss=0.1661, simple_loss=0.2525, pruned_loss=0.03989, over 7435.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2927, pruned_loss=0.06536, over 1619213.67 frames. ], batch size: 17, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:47:13,458 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152744.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:47:15,605 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6313, 1.5306, 2.1162, 1.3537, 1.1770, 2.0522, 0.3230, 1.2774], + device='cuda:1'), covar=tensor([0.1694, 0.1326, 0.0322, 0.1219, 0.2758, 0.0416, 0.2015, 0.1207], + device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0193, 0.0123, 0.0221, 0.0267, 0.0132, 0.0168, 0.0185], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 23:47:17,696 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:47:21,920 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1185, 1.4324, 1.7104, 1.3461, 0.9266, 1.4437, 1.7668, 1.4720], + device='cuda:1'), covar=tensor([0.0538, 0.1314, 0.1667, 0.1471, 0.0640, 0.1538, 0.0707, 0.0689], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0151, 0.0189, 0.0157, 0.0099, 0.0161, 0.0112, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 23:47:37,387 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.392e+02 2.877e+02 3.488e+02 7.359e+02, threshold=5.753e+02, percent-clipped=2.0 +2023-02-06 23:47:47,609 INFO [train.py:901] (1/4) Epoch 19, batch 7300, loss[loss=0.2241, simple_loss=0.3057, pruned_loss=0.07123, over 8568.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2919, pruned_loss=0.0651, over 1618018.35 frames. ], batch size: 31, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:47:57,327 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:48:21,883 INFO [train.py:901] (1/4) Epoch 19, batch 7350, loss[loss=0.2417, simple_loss=0.3046, pruned_loss=0.08943, over 7528.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2925, pruned_loss=0.06584, over 1613953.22 frames. ], batch size: 18, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:48:33,057 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5148, 1.4719, 1.8415, 1.2800, 1.1419, 1.8277, 0.2265, 1.1617], + device='cuda:1'), covar=tensor([0.1686, 0.1311, 0.0379, 0.1030, 0.2818, 0.0465, 0.2262, 0.1287], + device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0192, 0.0121, 0.0219, 0.0265, 0.0131, 0.0166, 0.0184], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 23:48:46,740 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 23:48:48,163 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.571e+02 3.070e+02 4.184e+02 8.940e+02, threshold=6.140e+02, percent-clipped=8.0 +2023-02-06 23:48:58,052 INFO [train.py:901] (1/4) Epoch 19, batch 7400, loss[loss=0.166, simple_loss=0.2419, pruned_loss=0.04511, over 7534.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2924, pruned_loss=0.06522, over 1616210.82 frames. ], batch size: 18, lr: 3.95e-03, grad_scale: 32.0 +2023-02-06 23:49:07,697 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 23:49:18,790 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152922.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:49:32,922 INFO [train.py:901] (1/4) Epoch 19, batch 7450, loss[loss=0.1886, simple_loss=0.2735, pruned_loss=0.05183, over 8249.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2912, pruned_loss=0.06447, over 1619560.74 frames. ], batch size: 24, lr: 3.95e-03, grad_scale: 32.0 +2023-02-06 23:49:33,699 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:49:38,521 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:49:39,907 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7057, 4.7337, 4.1922, 1.9157, 4.1327, 4.3712, 4.2111, 4.0756], + device='cuda:1'), covar=tensor([0.0681, 0.0470, 0.1042, 0.5064, 0.0827, 0.0709, 0.1115, 0.0599], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0429, 0.0429, 0.0535, 0.0419, 0.0430, 0.0413, 0.0374], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:49:44,018 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152958.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:49:46,567 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 23:49:58,932 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.506e+02 3.079e+02 4.075e+02 8.166e+02, threshold=6.159e+02, percent-clipped=5.0 +2023-02-06 23:50:01,986 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152983.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:50:08,330 INFO [train.py:901] (1/4) Epoch 19, batch 7500, loss[loss=0.2562, simple_loss=0.3313, pruned_loss=0.09056, over 8452.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2908, pruned_loss=0.06425, over 1619548.84 frames. ], batch size: 25, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:50:17,493 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:50:34,841 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153031.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:50:42,934 INFO [train.py:901] (1/4) Epoch 19, batch 7550, loss[loss=0.2403, simple_loss=0.3078, pruned_loss=0.08646, over 6803.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2906, pruned_loss=0.06443, over 1617733.82 frames. ], batch size: 71, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:50:53,915 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:50:58,703 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:51:06,106 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1378, 1.8610, 2.1053, 1.9965, 1.2622, 1.7953, 2.6783, 2.4781], + device='cuda:1'), covar=tensor([0.0413, 0.1158, 0.1546, 0.1237, 0.0532, 0.1376, 0.0491, 0.0545], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0100, 0.0161, 0.0113, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 23:51:08,489 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.431e+02 2.980e+02 3.688e+02 7.634e+02, threshold=5.960e+02, percent-clipped=2.0 +2023-02-06 23:51:10,130 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2463, 1.1979, 1.5384, 1.1477, 0.6901, 1.3084, 1.2111, 1.1720], + device='cuda:1'), covar=tensor([0.0576, 0.1339, 0.1735, 0.1508, 0.0604, 0.1540, 0.0724, 0.0645], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0100, 0.0161, 0.0113, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-06 23:51:14,082 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=153088.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:51:18,073 INFO [train.py:901] (1/4) Epoch 19, batch 7600, loss[loss=0.2123, simple_loss=0.2819, pruned_loss=0.07131, over 7436.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2913, pruned_loss=0.06484, over 1613202.31 frames. ], batch size: 17, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:51:47,782 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:51:53,104 INFO [train.py:901] (1/4) Epoch 19, batch 7650, loss[loss=0.2486, simple_loss=0.3118, pruned_loss=0.09265, over 7721.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2911, pruned_loss=0.06474, over 1613335.44 frames. ], batch size: 18, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:52:17,637 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153178.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:52:18,733 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.290e+02 2.780e+02 3.362e+02 7.829e+02, threshold=5.561e+02, percent-clipped=2.0 +2023-02-06 23:52:28,394 INFO [train.py:901] (1/4) Epoch 19, batch 7700, loss[loss=0.1905, simple_loss=0.2804, pruned_loss=0.05027, over 8255.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2921, pruned_loss=0.06541, over 1611319.38 frames. ], batch size: 24, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:52:35,485 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153203.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:52:35,509 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153203.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:52:52,969 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4691, 1.8598, 4.5298, 1.8526, 2.5023, 5.1360, 5.2256, 4.4963], + device='cuda:1'), covar=tensor([0.1062, 0.1632, 0.0244, 0.2053, 0.1161, 0.0168, 0.0332, 0.0498], + device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0320, 0.0286, 0.0314, 0.0303, 0.0264, 0.0407, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:52:57,461 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 23:53:03,343 INFO [train.py:901] (1/4) Epoch 19, batch 7750, loss[loss=0.2039, simple_loss=0.291, pruned_loss=0.05839, over 8493.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2918, pruned_loss=0.06534, over 1606160.11 frames. ], batch size: 28, lr: 3.94e-03, grad_scale: 16.0 +2023-02-06 23:53:28,914 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 2.456e+02 3.001e+02 3.725e+02 8.940e+02, threshold=6.003e+02, percent-clipped=11.0 +2023-02-06 23:53:37,743 INFO [train.py:901] (1/4) Epoch 19, batch 7800, loss[loss=0.2183, simple_loss=0.2957, pruned_loss=0.07047, over 8550.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2916, pruned_loss=0.06579, over 1606249.84 frames. ], batch size: 39, lr: 3.94e-03, grad_scale: 16.0 +2023-02-06 23:53:39,902 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153296.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:53:53,363 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153315.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:53:58,009 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:54:08,875 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2318, 4.2173, 3.8211, 2.0344, 3.7922, 3.7747, 3.7877, 3.6172], + device='cuda:1'), covar=tensor([0.0748, 0.0532, 0.1067, 0.4233, 0.0928, 0.0938, 0.1303, 0.0723], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0431, 0.0432, 0.0540, 0.0424, 0.0436, 0.0418, 0.0377], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:54:09,684 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153340.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:54:11,465 INFO [train.py:901] (1/4) Epoch 19, batch 7850, loss[loss=0.1856, simple_loss=0.2742, pruned_loss=0.04852, over 8195.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2918, pruned_loss=0.06545, over 1607235.88 frames. ], batch size: 23, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:54:14,369 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:54:30,516 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-06 23:54:36,623 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.457e+02 2.874e+02 3.581e+02 1.670e+03, threshold=5.749e+02, percent-clipped=9.0 +2023-02-06 23:54:44,305 INFO [train.py:901] (1/4) Epoch 19, batch 7900, loss[loss=0.2117, simple_loss=0.3015, pruned_loss=0.061, over 8197.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2921, pruned_loss=0.06526, over 1610560.82 frames. ], batch size: 23, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:55:08,825 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-06 23:55:15,513 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153439.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:55:17,980 INFO [train.py:901] (1/4) Epoch 19, batch 7950, loss[loss=0.2237, simple_loss=0.305, pruned_loss=0.07115, over 8285.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2926, pruned_loss=0.06532, over 1615823.34 frames. ], batch size: 48, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:55:28,831 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153459.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:55:41,874 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=153479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:55:43,178 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 2.434e+02 3.034e+02 3.983e+02 8.510e+02, threshold=6.068e+02, percent-clipped=6.0 +2023-02-06 23:55:45,323 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153484.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:55:47,949 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9760, 1.6232, 1.4764, 1.6122, 1.3326, 1.3004, 1.2952, 1.3599], + device='cuda:1'), covar=tensor([0.1144, 0.0455, 0.1147, 0.0538, 0.0735, 0.1375, 0.0856, 0.0739], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0236, 0.0330, 0.0305, 0.0298, 0.0332, 0.0343, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-06 23:55:51,002 INFO [train.py:901] (1/4) Epoch 19, batch 8000, loss[loss=0.2264, simple_loss=0.3182, pruned_loss=0.06728, over 8331.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2931, pruned_loss=0.06558, over 1612006.35 frames. ], batch size: 26, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:56:10,304 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4847, 1.5062, 1.8661, 1.3183, 1.2205, 1.8428, 0.2761, 1.2324], + device='cuda:1'), covar=tensor([0.1703, 0.1281, 0.0358, 0.0949, 0.2832, 0.0408, 0.2093, 0.1220], + device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0194, 0.0124, 0.0222, 0.0270, 0.0133, 0.0169, 0.0186], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-06 23:56:25,155 INFO [train.py:901] (1/4) Epoch 19, batch 8050, loss[loss=0.188, simple_loss=0.25, pruned_loss=0.06303, over 7536.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2928, pruned_loss=0.06601, over 1603259.36 frames. ], batch size: 18, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:56:59,001 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 23:57:04,933 INFO [train.py:901] (1/4) Epoch 20, batch 0, loss[loss=0.2719, simple_loss=0.3423, pruned_loss=0.1007, over 8388.00 frames. ], tot_loss[loss=0.2719, simple_loss=0.3423, pruned_loss=0.1007, over 8388.00 frames. ], batch size: 48, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:57:04,933 INFO [train.py:926] (1/4) Computing validation loss +2023-02-06 23:57:16,943 INFO [train.py:935] (1/4) Epoch 20, validation: loss=0.1757, simple_loss=0.276, pruned_loss=0.03766, over 944034.00 frames. +2023-02-06 23:57:16,946 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-06 23:57:20,456 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.577e+02 3.496e+02 4.495e+02 1.164e+03, threshold=6.992e+02, percent-clipped=12.0 +2023-02-06 23:57:29,441 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:57:31,313 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 23:57:51,315 INFO [train.py:901] (1/4) Epoch 20, batch 50, loss[loss=0.208, simple_loss=0.2824, pruned_loss=0.06682, over 7796.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2967, pruned_loss=0.06733, over 365647.95 frames. ], batch size: 19, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:58:01,101 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=153640.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:58:06,575 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 23:58:27,836 INFO [train.py:901] (1/4) Epoch 20, batch 100, loss[loss=0.2232, simple_loss=0.2994, pruned_loss=0.07355, over 7922.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2953, pruned_loss=0.06562, over 647534.65 frames. ], batch size: 20, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:58:29,238 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 23:58:31,349 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.446e+02 2.844e+02 3.351e+02 7.473e+02, threshold=5.688e+02, percent-clipped=2.0 +2023-02-06 23:58:58,569 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.5990, 1.7129, 5.7461, 2.2379, 5.1285, 4.8869, 5.2775, 5.1902], + device='cuda:1'), covar=tensor([0.0524, 0.4918, 0.0423, 0.3752, 0.1014, 0.0866, 0.0514, 0.0532], + device='cuda:1'), in_proj_covar=tensor([0.0607, 0.0639, 0.0679, 0.0613, 0.0691, 0.0595, 0.0593, 0.0655], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:59:03,132 INFO [train.py:901] (1/4) Epoch 20, batch 150, loss[loss=0.2531, simple_loss=0.3257, pruned_loss=0.09021, over 8589.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2931, pruned_loss=0.06452, over 862779.74 frames. ], batch size: 31, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:59:08,801 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3795, 1.5681, 4.5548, 1.7888, 4.0439, 3.8052, 4.1269, 4.0124], + device='cuda:1'), covar=tensor([0.0523, 0.4580, 0.0508, 0.3773, 0.1039, 0.0941, 0.0524, 0.0650], + device='cuda:1'), in_proj_covar=tensor([0.0607, 0.0639, 0.0679, 0.0613, 0.0691, 0.0595, 0.0593, 0.0656], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-06 23:59:23,338 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:59:39,296 INFO [train.py:901] (1/4) Epoch 20, batch 200, loss[loss=0.2027, simple_loss=0.2856, pruned_loss=0.05992, over 8513.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.291, pruned_loss=0.06387, over 1030295.22 frames. ], batch size: 26, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:59:42,519 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 2.177e+02 2.784e+02 3.416e+02 8.818e+02, threshold=5.569e+02, percent-clipped=1.0 +2023-02-06 23:59:43,903 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=153783.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:59:51,028 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.91 vs. limit=5.0 +2023-02-06 23:59:58,792 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 00:00:15,029 INFO [train.py:901] (1/4) Epoch 20, batch 250, loss[loss=0.231, simple_loss=0.3097, pruned_loss=0.0761, over 8404.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2914, pruned_loss=0.06414, over 1160146.55 frames. ], batch size: 49, lr: 3.84e-03, grad_scale: 8.0 +2023-02-07 00:00:26,546 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 00:00:31,635 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153850.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:00:34,736 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 00:00:48,276 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153875.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:00:48,774 INFO [train.py:901] (1/4) Epoch 20, batch 300, loss[loss=0.1827, simple_loss=0.2655, pruned_loss=0.04999, over 8094.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2921, pruned_loss=0.06479, over 1263757.76 frames. ], batch size: 21, lr: 3.84e-03, grad_scale: 8.0 +2023-02-07 00:00:51,993 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.425e+02 2.846e+02 3.739e+02 1.062e+03, threshold=5.691e+02, percent-clipped=2.0 +2023-02-07 00:01:05,166 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153898.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:01:24,551 INFO [train.py:901] (1/4) Epoch 20, batch 350, loss[loss=0.2387, simple_loss=0.3013, pruned_loss=0.08803, over 7977.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2929, pruned_loss=0.0644, over 1345264.79 frames. ], batch size: 21, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:01:35,759 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153941.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:01:54,117 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5685, 2.3216, 1.6686, 2.2682, 2.1070, 1.3365, 2.0706, 2.2168], + device='cuda:1'), covar=tensor([0.1378, 0.0448, 0.1329, 0.0609, 0.0804, 0.1701, 0.0989, 0.0824], + device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0233, 0.0329, 0.0303, 0.0297, 0.0331, 0.0342, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 00:01:54,352 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-07 00:01:59,294 INFO [train.py:901] (1/4) Epoch 20, batch 400, loss[loss=0.2182, simple_loss=0.302, pruned_loss=0.06726, over 8493.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2932, pruned_loss=0.06515, over 1403070.13 frames. ], batch size: 29, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:02:02,809 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 2.483e+02 2.937e+02 3.652e+02 9.410e+02, threshold=5.874e+02, percent-clipped=4.0 +2023-02-07 00:02:25,628 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154011.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:02:36,314 INFO [train.py:901] (1/4) Epoch 20, batch 450, loss[loss=0.2372, simple_loss=0.3159, pruned_loss=0.07926, over 8365.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2927, pruned_loss=0.0652, over 1450270.69 frames. ], batch size: 49, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:02:44,062 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154036.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:03:05,764 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154067.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:03:11,796 INFO [train.py:901] (1/4) Epoch 20, batch 500, loss[loss=0.2286, simple_loss=0.3282, pruned_loss=0.06457, over 8493.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2912, pruned_loss=0.065, over 1478922.61 frames. ], batch size: 26, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:03:13,450 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5022, 2.4203, 3.3192, 2.5618, 3.0961, 2.5631, 2.3786, 1.8285], + device='cuda:1'), covar=tensor([0.5214, 0.4948, 0.1766, 0.4010, 0.2632, 0.2814, 0.1754, 0.5436], + device='cuda:1'), in_proj_covar=tensor([0.0931, 0.0961, 0.0789, 0.0926, 0.0980, 0.0871, 0.0736, 0.0815], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 00:03:15,232 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 2.274e+02 2.685e+02 3.204e+02 7.760e+02, threshold=5.371e+02, percent-clipped=3.0 +2023-02-07 00:03:24,407 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154094.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:03:29,960 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154102.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:03:46,380 INFO [train.py:901] (1/4) Epoch 20, batch 550, loss[loss=0.199, simple_loss=0.2702, pruned_loss=0.06389, over 8089.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2912, pruned_loss=0.06502, over 1512258.83 frames. ], batch size: 21, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:04:07,824 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154154.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:04:23,281 INFO [train.py:901] (1/4) Epoch 20, batch 600, loss[loss=0.1988, simple_loss=0.2744, pruned_loss=0.06157, over 7648.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2918, pruned_loss=0.06571, over 1531439.49 frames. ], batch size: 19, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:04:25,541 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154179.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:04:26,654 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.477e+02 2.962e+02 3.836e+02 8.919e+02, threshold=5.925e+02, percent-clipped=6.0 +2023-02-07 00:04:45,284 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 00:04:57,562 INFO [train.py:901] (1/4) Epoch 20, batch 650, loss[loss=0.2021, simple_loss=0.2895, pruned_loss=0.05735, over 8450.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2919, pruned_loss=0.06532, over 1554150.20 frames. ], batch size: 27, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:05:06,573 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:05:34,077 INFO [train.py:901] (1/4) Epoch 20, batch 700, loss[loss=0.2374, simple_loss=0.3196, pruned_loss=0.07759, over 8458.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2913, pruned_loss=0.06475, over 1570884.55 frames. ], batch size: 25, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:05:37,460 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.357e+02 2.958e+02 3.586e+02 6.466e+02, threshold=5.915e+02, percent-clipped=2.0 +2023-02-07 00:05:40,247 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154285.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:05:54,062 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154304.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:05:57,554 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154309.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:06:08,850 INFO [train.py:901] (1/4) Epoch 20, batch 750, loss[loss=0.1923, simple_loss=0.2788, pruned_loss=0.05287, over 7800.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2908, pruned_loss=0.06433, over 1582096.55 frames. ], batch size: 20, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:06:28,503 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154355.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:06:31,837 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1893, 2.1247, 1.6089, 1.8818, 1.7954, 1.2864, 1.6613, 1.6305], + device='cuda:1'), covar=tensor([0.1324, 0.0419, 0.1134, 0.0524, 0.0726, 0.1536, 0.0868, 0.0874], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0236, 0.0330, 0.0305, 0.0300, 0.0334, 0.0344, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 00:06:33,651 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 00:06:43,003 INFO [train.py:901] (1/4) Epoch 20, batch 800, loss[loss=0.1927, simple_loss=0.2845, pruned_loss=0.05045, over 8473.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2905, pruned_loss=0.06424, over 1592869.52 frames. ], batch size: 27, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:06:43,011 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 00:06:47,164 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 2.441e+02 3.052e+02 3.711e+02 8.675e+02, threshold=6.104e+02, percent-clipped=3.0 +2023-02-07 00:07:01,167 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154400.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:07:07,113 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7216, 2.0193, 2.2243, 1.5691, 2.3787, 1.5163, 0.7646, 1.9103], + device='cuda:1'), covar=tensor([0.0540, 0.0290, 0.0196, 0.0429, 0.0286, 0.0670, 0.0727, 0.0259], + device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0377, 0.0332, 0.0436, 0.0362, 0.0523, 0.0381, 0.0404], + device='cuda:1'), out_proj_covar=tensor([1.1821e-04, 9.9056e-05, 8.7847e-05, 1.1573e-04, 9.5808e-05, 1.4897e-04, + 1.0306e-04, 1.0783e-04], device='cuda:1') +2023-02-07 00:07:08,303 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154411.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:07:19,172 INFO [train.py:901] (1/4) Epoch 20, batch 850, loss[loss=0.2054, simple_loss=0.2884, pruned_loss=0.06122, over 8109.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2907, pruned_loss=0.06374, over 1601320.13 frames. ], batch size: 23, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:07:27,214 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154438.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:07:32,659 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154446.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:07:52,875 INFO [train.py:901] (1/4) Epoch 20, batch 900, loss[loss=0.1931, simple_loss=0.2721, pruned_loss=0.05703, over 7507.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2898, pruned_loss=0.0636, over 1601853.45 frames. ], batch size: 18, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:07:56,214 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 2.439e+02 2.923e+02 3.686e+02 1.072e+03, threshold=5.846e+02, percent-clipped=2.0 +2023-02-07 00:08:00,439 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6621, 1.4579, 1.7241, 1.3547, 0.9076, 1.4469, 1.5432, 1.5413], + device='cuda:1'), covar=tensor([0.0568, 0.1247, 0.1584, 0.1479, 0.0597, 0.1454, 0.0667, 0.0611], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0100, 0.0161, 0.0113, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 00:08:04,405 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154492.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:08:29,161 INFO [train.py:901] (1/4) Epoch 20, batch 950, loss[loss=0.3014, simple_loss=0.3502, pruned_loss=0.1263, over 6712.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.291, pruned_loss=0.06429, over 1601911.68 frames. ], batch size: 71, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:08:29,375 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154526.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:08:30,901 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.26 vs. limit=5.0 +2023-02-07 00:08:48,724 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154553.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:08:54,157 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154561.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:09:01,513 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 00:09:04,232 INFO [train.py:901] (1/4) Epoch 20, batch 1000, loss[loss=0.2232, simple_loss=0.3058, pruned_loss=0.07032, over 8439.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2907, pruned_loss=0.06408, over 1600725.74 frames. ], batch size: 27, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:09:07,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.511e+02 3.044e+02 3.807e+02 8.767e+02, threshold=6.087e+02, percent-clipped=2.0 +2023-02-07 00:09:08,972 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154583.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:09:27,672 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-07 00:09:35,151 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 00:09:38,954 INFO [train.py:901] (1/4) Epoch 20, batch 1050, loss[loss=0.1608, simple_loss=0.2485, pruned_loss=0.03653, over 7795.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2897, pruned_loss=0.06385, over 1599691.16 frames. ], batch size: 20, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:09:41,228 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5773, 1.8213, 1.9690, 1.3222, 2.1308, 1.4209, 0.5623, 1.7914], + device='cuda:1'), covar=tensor([0.0600, 0.0359, 0.0303, 0.0569, 0.0415, 0.0836, 0.0815, 0.0317], + device='cuda:1'), in_proj_covar=tensor([0.0436, 0.0377, 0.0332, 0.0436, 0.0362, 0.0523, 0.0381, 0.0403], + device='cuda:1'), out_proj_covar=tensor([1.1771e-04, 9.9159e-05, 8.7781e-05, 1.1574e-04, 9.5813e-05, 1.4894e-04, + 1.0321e-04, 1.0761e-04], device='cuda:1') +2023-02-07 00:09:49,448 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 00:09:53,676 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154646.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:09:54,883 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154648.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:09:59,104 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154653.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:10:01,446 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154656.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:10:14,713 INFO [train.py:901] (1/4) Epoch 20, batch 1100, loss[loss=0.2115, simple_loss=0.2901, pruned_loss=0.06646, over 8651.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.289, pruned_loss=0.064, over 1598274.88 frames. ], batch size: 34, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:10:18,094 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.486e+02 3.103e+02 3.988e+02 8.246e+02, threshold=6.206e+02, percent-clipped=6.0 +2023-02-07 00:10:18,321 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154681.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:10:29,761 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154698.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:10:30,329 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154699.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:10:48,875 INFO [train.py:901] (1/4) Epoch 20, batch 1150, loss[loss=0.1856, simple_loss=0.2785, pruned_loss=0.0463, over 7809.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2901, pruned_loss=0.06438, over 1603584.76 frames. ], batch size: 20, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:10:57,842 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154738.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:10:59,088 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 00:11:16,280 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154763.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:11:19,778 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:11:25,027 INFO [train.py:901] (1/4) Epoch 20, batch 1200, loss[loss=0.2169, simple_loss=0.2883, pruned_loss=0.07281, over 7648.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2891, pruned_loss=0.06363, over 1605300.36 frames. ], batch size: 19, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:11:28,373 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.412e+02 2.746e+02 3.577e+02 9.067e+02, threshold=5.492e+02, percent-clipped=2.0 +2023-02-07 00:11:29,309 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154782.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:11:46,342 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154807.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:11:47,765 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154809.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:11:51,159 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154814.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:11:53,285 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154817.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:11:59,125 INFO [train.py:901] (1/4) Epoch 20, batch 1250, loss[loss=0.1947, simple_loss=0.2803, pruned_loss=0.05455, over 8247.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2907, pruned_loss=0.0642, over 1607021.65 frames. ], batch size: 24, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:12:05,337 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154834.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:12:06,468 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154836.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:12:11,367 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154842.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:12:26,513 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 00:12:34,977 INFO [train.py:901] (1/4) Epoch 20, batch 1300, loss[loss=0.2069, simple_loss=0.2752, pruned_loss=0.06924, over 7802.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2892, pruned_loss=0.0638, over 1606940.59 frames. ], batch size: 19, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:12:38,316 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.433e+02 3.191e+02 3.995e+02 7.235e+02, threshold=6.381e+02, percent-clipped=6.0 +2023-02-07 00:12:51,205 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7165, 1.9965, 2.1600, 1.3761, 2.3550, 1.5715, 0.7610, 1.9664], + device='cuda:1'), covar=tensor([0.0649, 0.0371, 0.0290, 0.0611, 0.0374, 0.0914, 0.0868, 0.0327], + device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0383, 0.0335, 0.0442, 0.0367, 0.0528, 0.0387, 0.0408], + device='cuda:1'), out_proj_covar=tensor([1.1908e-04, 1.0093e-04, 8.8555e-05, 1.1715e-04, 9.7293e-05, 1.5050e-04, + 1.0489e-04, 1.0898e-04], device='cuda:1') +2023-02-07 00:12:52,158 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 00:13:05,460 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7629, 1.8183, 2.3366, 1.5873, 1.3525, 2.2984, 0.4782, 1.3628], + device='cuda:1'), covar=tensor([0.1885, 0.1237, 0.0337, 0.1324, 0.3010, 0.0417, 0.2244, 0.1476], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0192, 0.0123, 0.0220, 0.0268, 0.0133, 0.0169, 0.0187], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 00:13:09,371 INFO [train.py:901] (1/4) Epoch 20, batch 1350, loss[loss=0.17, simple_loss=0.2499, pruned_loss=0.04505, over 7429.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2894, pruned_loss=0.06378, over 1608087.88 frames. ], batch size: 17, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:13:27,091 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154951.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:13:29,048 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154954.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:13:40,767 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2246, 1.2611, 1.4711, 1.2467, 0.7251, 1.2599, 1.1643, 1.0361], + device='cuda:1'), covar=tensor([0.0560, 0.1241, 0.1634, 0.1359, 0.0566, 0.1486, 0.0716, 0.0686], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0157, 0.0100, 0.0160, 0.0112, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 00:13:44,756 INFO [train.py:901] (1/4) Epoch 20, batch 1400, loss[loss=0.1826, simple_loss=0.2625, pruned_loss=0.05138, over 7544.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2905, pruned_loss=0.0643, over 1610583.90 frames. ], batch size: 18, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:13:47,812 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154979.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:13:48,965 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.419e+02 2.969e+02 3.620e+02 8.609e+02, threshold=5.938e+02, percent-clipped=3.0 +2023-02-07 00:13:55,298 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:13:58,170 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154994.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:16,374 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155019.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:19,578 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155024.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:20,734 INFO [train.py:901] (1/4) Epoch 20, batch 1450, loss[loss=0.1888, simple_loss=0.2858, pruned_loss=0.04591, over 8717.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.291, pruned_loss=0.06465, over 1611076.63 frames. ], batch size: 34, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:14:29,172 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 00:14:33,302 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155044.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:36,493 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155049.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:50,887 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155070.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:14:55,251 INFO [train.py:901] (1/4) Epoch 20, batch 1500, loss[loss=0.19, simple_loss=0.2828, pruned_loss=0.04856, over 8252.00 frames. ], tot_loss[loss=0.21, simple_loss=0.291, pruned_loss=0.06453, over 1615834.92 frames. ], batch size: 24, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:14:58,584 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.504e+02 2.482e+02 3.072e+02 3.822e+02 6.990e+02, threshold=6.143e+02, percent-clipped=2.0 +2023-02-07 00:14:59,297 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155082.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:15:02,788 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155087.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:15:09,002 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155095.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:15:15,565 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155105.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:15:30,481 INFO [train.py:901] (1/4) Epoch 20, batch 1550, loss[loss=0.2231, simple_loss=0.2976, pruned_loss=0.07428, over 8290.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2903, pruned_loss=0.064, over 1615127.21 frames. ], batch size: 23, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:15:37,001 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1980, 1.9838, 2.6763, 2.2052, 2.6984, 2.2388, 1.9575, 1.4023], + device='cuda:1'), covar=tensor([0.5537, 0.5039, 0.2091, 0.3956, 0.2643, 0.3110, 0.2002, 0.5668], + device='cuda:1'), in_proj_covar=tensor([0.0930, 0.0960, 0.0787, 0.0925, 0.0981, 0.0874, 0.0736, 0.0815], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 00:15:44,485 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 00:16:04,723 INFO [train.py:901] (1/4) Epoch 20, batch 1600, loss[loss=0.1986, simple_loss=0.2832, pruned_loss=0.05697, over 8100.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2898, pruned_loss=0.06358, over 1615509.52 frames. ], batch size: 23, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:16:08,760 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.295e+02 2.863e+02 3.431e+02 6.352e+02, threshold=5.726e+02, percent-clipped=1.0 +2023-02-07 00:16:20,595 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155197.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:16:27,182 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155207.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:16:31,900 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5756, 1.8428, 2.0006, 1.4923, 2.1642, 1.4820, 0.6204, 1.8340], + device='cuda:1'), covar=tensor([0.0525, 0.0332, 0.0233, 0.0426, 0.0340, 0.0799, 0.0822, 0.0257], + device='cuda:1'), in_proj_covar=tensor([0.0440, 0.0382, 0.0336, 0.0440, 0.0368, 0.0527, 0.0387, 0.0407], + device='cuda:1'), out_proj_covar=tensor([1.1892e-04, 1.0037e-04, 8.8691e-05, 1.1646e-04, 9.7576e-05, 1.5007e-04, + 1.0492e-04, 1.0855e-04], device='cuda:1') +2023-02-07 00:16:40,723 INFO [train.py:901] (1/4) Epoch 20, batch 1650, loss[loss=0.1903, simple_loss=0.2805, pruned_loss=0.05008, over 7927.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2894, pruned_loss=0.06363, over 1611556.36 frames. ], batch size: 20, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:16:45,155 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155232.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:17:15,943 INFO [train.py:901] (1/4) Epoch 20, batch 1700, loss[loss=0.2121, simple_loss=0.299, pruned_loss=0.06263, over 8462.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2904, pruned_loss=0.06419, over 1614644.87 frames. ], batch size: 27, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:17:19,373 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.383e+02 2.759e+02 3.259e+02 7.427e+02, threshold=5.517e+02, percent-clipped=3.0 +2023-02-07 00:17:51,289 INFO [train.py:901] (1/4) Epoch 20, batch 1750, loss[loss=0.1776, simple_loss=0.2648, pruned_loss=0.04524, over 7986.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2917, pruned_loss=0.06464, over 1617525.39 frames. ], batch size: 21, lr: 3.82e-03, grad_scale: 16.0 +2023-02-07 00:18:00,381 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155338.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:18:17,135 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155361.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:18:27,001 INFO [train.py:901] (1/4) Epoch 20, batch 1800, loss[loss=0.2508, simple_loss=0.3297, pruned_loss=0.08596, over 8453.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2911, pruned_loss=0.06455, over 1614788.58 frames. ], batch size: 27, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:18:31,088 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.586e+02 2.965e+02 3.772e+02 7.314e+02, threshold=5.929e+02, percent-clipped=8.0 +2023-02-07 00:18:34,026 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155386.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:18:40,877 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 00:18:51,794 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.87 vs. limit=5.0 +2023-02-07 00:19:01,120 INFO [train.py:901] (1/4) Epoch 20, batch 1850, loss[loss=0.1705, simple_loss=0.2738, pruned_loss=0.03364, over 8351.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2914, pruned_loss=0.06458, over 1614770.14 frames. ], batch size: 24, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:19:04,533 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155431.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:20,145 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155453.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:20,188 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155453.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:36,712 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155475.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:37,253 INFO [train.py:901] (1/4) Epoch 20, batch 1900, loss[loss=0.1995, simple_loss=0.2839, pruned_loss=0.05754, over 7927.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2907, pruned_loss=0.06409, over 1613865.01 frames. ], batch size: 20, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:19:38,725 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2532, 1.1530, 3.3355, 1.0435, 2.9392, 2.7591, 3.0387, 2.8856], + device='cuda:1'), covar=tensor([0.0711, 0.4525, 0.0759, 0.4167, 0.1297, 0.1134, 0.0720, 0.0931], + device='cuda:1'), in_proj_covar=tensor([0.0602, 0.0624, 0.0673, 0.0604, 0.0684, 0.0588, 0.0587, 0.0656], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 00:19:38,784 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155478.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:41,336 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.441e+02 2.899e+02 3.473e+02 6.405e+02, threshold=5.799e+02, percent-clipped=1.0 +2023-02-07 00:20:11,851 INFO [train.py:901] (1/4) Epoch 20, batch 1950, loss[loss=0.2144, simple_loss=0.2894, pruned_loss=0.06969, over 7974.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2901, pruned_loss=0.06331, over 1615693.08 frames. ], batch size: 21, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:20:13,303 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 00:20:26,412 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155546.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:20:26,922 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 00:20:29,836 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1312, 2.4023, 1.9570, 3.0071, 1.4138, 1.7761, 2.2020, 2.3572], + device='cuda:1'), covar=tensor([0.0738, 0.0810, 0.0872, 0.0351, 0.1144, 0.1285, 0.0926, 0.0804], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0198, 0.0248, 0.0213, 0.0205, 0.0250, 0.0252, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 00:20:46,965 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 00:20:47,677 INFO [train.py:901] (1/4) Epoch 20, batch 2000, loss[loss=0.1899, simple_loss=0.2767, pruned_loss=0.05158, over 8472.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2897, pruned_loss=0.06324, over 1610331.55 frames. ], batch size: 25, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:20:51,752 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.363e+02 2.911e+02 3.881e+02 1.027e+03, threshold=5.822e+02, percent-clipped=2.0 +2023-02-07 00:21:20,956 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155623.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:21:22,875 INFO [train.py:901] (1/4) Epoch 20, batch 2050, loss[loss=0.222, simple_loss=0.3063, pruned_loss=0.06885, over 8455.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2905, pruned_loss=0.06393, over 1611558.46 frames. ], batch size: 27, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:21:56,395 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 00:21:57,539 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155675.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:21:58,087 INFO [train.py:901] (1/4) Epoch 20, batch 2100, loss[loss=0.2334, simple_loss=0.3123, pruned_loss=0.07727, over 8503.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2913, pruned_loss=0.0645, over 1613042.39 frames. ], batch size: 26, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:22:02,103 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.564e+02 2.968e+02 3.686e+02 8.256e+02, threshold=5.935e+02, percent-clipped=7.0 +2023-02-07 00:22:22,169 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155709.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:22:33,481 INFO [train.py:901] (1/4) Epoch 20, batch 2150, loss[loss=0.1954, simple_loss=0.2737, pruned_loss=0.05857, over 8236.00 frames. ], tot_loss[loss=0.211, simple_loss=0.292, pruned_loss=0.06499, over 1614774.34 frames. ], batch size: 22, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:22:39,028 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155734.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:23:07,320 INFO [train.py:901] (1/4) Epoch 20, batch 2200, loss[loss=0.207, simple_loss=0.2932, pruned_loss=0.06037, over 8505.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2909, pruned_loss=0.06412, over 1618311.61 frames. ], batch size: 28, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:23:12,096 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.799e+02 2.519e+02 2.939e+02 3.787e+02 7.175e+02, threshold=5.878e+02, percent-clipped=4.0 +2023-02-07 00:23:26,038 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155802.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:23:26,117 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155802.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:23:38,412 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155819.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:23:43,079 INFO [train.py:901] (1/4) Epoch 20, batch 2250, loss[loss=0.1903, simple_loss=0.2814, pruned_loss=0.04964, over 8593.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2905, pruned_loss=0.06374, over 1618798.53 frames. ], batch size: 34, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:23:44,828 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155827.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:24:17,875 INFO [train.py:901] (1/4) Epoch 20, batch 2300, loss[loss=0.2085, simple_loss=0.2872, pruned_loss=0.0649, over 7543.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2908, pruned_loss=0.06406, over 1614426.30 frames. ], batch size: 18, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:24:21,984 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.500e+02 2.966e+02 3.753e+02 6.656e+02, threshold=5.933e+02, percent-clipped=3.0 +2023-02-07 00:24:54,615 INFO [train.py:901] (1/4) Epoch 20, batch 2350, loss[loss=0.2056, simple_loss=0.2942, pruned_loss=0.05845, over 8026.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2903, pruned_loss=0.06327, over 1614346.05 frames. ], batch size: 22, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:25:00,025 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155934.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:25:23,233 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155967.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:25:29,309 INFO [train.py:901] (1/4) Epoch 20, batch 2400, loss[loss=0.1986, simple_loss=0.2697, pruned_loss=0.06375, over 8134.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.292, pruned_loss=0.06455, over 1614398.73 frames. ], batch size: 22, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:25:33,222 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.377e+02 2.729e+02 3.502e+02 6.388e+02, threshold=5.458e+02, percent-clipped=1.0 +2023-02-07 00:26:00,992 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=156019.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:26:05,604 INFO [train.py:901] (1/4) Epoch 20, batch 2450, loss[loss=0.2027, simple_loss=0.2961, pruned_loss=0.05462, over 8249.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2923, pruned_loss=0.06466, over 1616101.93 frames. ], batch size: 24, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:26:35,053 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2499, 2.5378, 2.9008, 1.6863, 3.2386, 1.9453, 1.4798, 2.2295], + device='cuda:1'), covar=tensor([0.0687, 0.0406, 0.0288, 0.0707, 0.0359, 0.0781, 0.0876, 0.0477], + device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0380, 0.0334, 0.0437, 0.0366, 0.0525, 0.0386, 0.0407], + device='cuda:1'), out_proj_covar=tensor([1.1893e-04, 1.0001e-04, 8.8036e-05, 1.1564e-04, 9.6854e-05, 1.4930e-04, + 1.0449e-04, 1.0845e-04], device='cuda:1') +2023-02-07 00:26:40,969 INFO [train.py:901] (1/4) Epoch 20, batch 2500, loss[loss=0.2096, simple_loss=0.3017, pruned_loss=0.05877, over 8139.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2923, pruned_loss=0.06478, over 1610760.63 frames. ], batch size: 22, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:26:45,023 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 2.463e+02 3.105e+02 3.826e+02 1.382e+03, threshold=6.210e+02, percent-clipped=11.0 +2023-02-07 00:26:45,217 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156082.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:26:49,183 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156088.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:26:53,948 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3340, 1.2872, 2.3642, 1.3092, 2.1674, 2.5051, 2.6644, 2.1441], + device='cuda:1'), covar=tensor([0.1171, 0.1445, 0.0495, 0.2139, 0.0796, 0.0404, 0.0738, 0.0717], + device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0318, 0.0284, 0.0311, 0.0300, 0.0260, 0.0405, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 00:27:15,375 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4837, 2.4850, 1.7734, 2.1848, 2.0141, 1.5101, 1.8904, 2.0873], + device='cuda:1'), covar=tensor([0.1421, 0.0425, 0.1189, 0.0619, 0.0688, 0.1570, 0.1007, 0.0957], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0235, 0.0330, 0.0306, 0.0298, 0.0335, 0.0342, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 00:27:15,812 INFO [train.py:901] (1/4) Epoch 20, batch 2550, loss[loss=0.2321, simple_loss=0.3087, pruned_loss=0.07777, over 8026.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2927, pruned_loss=0.06504, over 1612235.34 frames. ], batch size: 22, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:27:21,374 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156134.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:27:29,846 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=156146.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:27:44,239 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5752, 2.6659, 2.0573, 2.4256, 2.2987, 1.8294, 2.0572, 2.2605], + device='cuda:1'), covar=tensor([0.1538, 0.0415, 0.1066, 0.0606, 0.0674, 0.1329, 0.1025, 0.0955], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0236, 0.0331, 0.0306, 0.0298, 0.0336, 0.0343, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 00:27:50,759 INFO [train.py:901] (1/4) Epoch 20, batch 2600, loss[loss=0.2284, simple_loss=0.3047, pruned_loss=0.07604, over 8474.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2915, pruned_loss=0.06436, over 1613989.45 frames. ], batch size: 25, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:27:54,662 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.376e+02 3.118e+02 3.808e+02 9.704e+02, threshold=6.236e+02, percent-clipped=5.0 +2023-02-07 00:28:00,362 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156190.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:28:16,947 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-07 00:28:17,477 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156215.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:28:24,598 INFO [train.py:901] (1/4) Epoch 20, batch 2650, loss[loss=0.2247, simple_loss=0.3094, pruned_loss=0.06997, over 8478.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2917, pruned_loss=0.0645, over 1616061.02 frames. ], batch size: 27, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:28:27,928 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8419, 3.7915, 3.4267, 1.7545, 3.3689, 3.4491, 3.3968, 3.2011], + device='cuda:1'), covar=tensor([0.0855, 0.0647, 0.1108, 0.4734, 0.0903, 0.0966, 0.1494, 0.0961], + device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0425, 0.0429, 0.0526, 0.0419, 0.0431, 0.0417, 0.0372], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:28:28,689 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7588, 1.6700, 1.7854, 1.5766, 1.0215, 1.5976, 2.0586, 1.8841], + device='cuda:1'), covar=tensor([0.0476, 0.1234, 0.1708, 0.1392, 0.0607, 0.1514, 0.0642, 0.0604], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0151, 0.0190, 0.0158, 0.0100, 0.0161, 0.0112, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 00:28:30,656 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156234.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:28:49,174 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:29:00,117 INFO [train.py:901] (1/4) Epoch 20, batch 2700, loss[loss=0.2186, simple_loss=0.2986, pruned_loss=0.06933, over 8484.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2919, pruned_loss=0.0647, over 1616604.41 frames. ], batch size: 28, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:29:02,512 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-07 00:29:03,627 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7061, 2.5655, 1.8222, 2.3244, 2.3411, 1.6112, 2.1648, 2.2330], + device='cuda:1'), covar=tensor([0.1335, 0.0367, 0.1093, 0.0564, 0.0580, 0.1419, 0.0866, 0.0902], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0236, 0.0333, 0.0307, 0.0300, 0.0338, 0.0344, 0.0319], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 00:29:04,092 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.401e+02 3.078e+02 3.829e+02 8.557e+02, threshold=6.156e+02, percent-clipped=4.0 +2023-02-07 00:29:23,253 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156308.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:29:35,129 INFO [train.py:901] (1/4) Epoch 20, batch 2750, loss[loss=0.2216, simple_loss=0.2912, pruned_loss=0.07598, over 7651.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2923, pruned_loss=0.06506, over 1615259.43 frames. ], batch size: 19, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:29:40,761 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7592, 2.6462, 1.8529, 2.3246, 2.3256, 1.6020, 2.1274, 2.2316], + device='cuda:1'), covar=tensor([0.1517, 0.0414, 0.1227, 0.0717, 0.0761, 0.1543, 0.1120, 0.1168], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0236, 0.0332, 0.0307, 0.0299, 0.0337, 0.0343, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 00:29:43,522 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156338.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:29:59,920 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 00:30:01,796 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156363.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:30:10,370 INFO [train.py:901] (1/4) Epoch 20, batch 2800, loss[loss=0.2697, simple_loss=0.3279, pruned_loss=0.1057, over 8535.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2926, pruned_loss=0.06576, over 1614959.56 frames. ], batch size: 49, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:30:15,863 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 2.534e+02 2.983e+02 3.648e+02 6.974e+02, threshold=5.966e+02, percent-clipped=1.0 +2023-02-07 00:30:20,855 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156390.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:30:38,823 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156415.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:30:46,250 INFO [train.py:901] (1/4) Epoch 20, batch 2850, loss[loss=0.1852, simple_loss=0.2774, pruned_loss=0.04651, over 8290.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2927, pruned_loss=0.06543, over 1617405.54 frames. ], batch size: 23, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:30:50,404 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=156432.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:31:10,265 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-07 00:31:20,791 INFO [train.py:901] (1/4) Epoch 20, batch 2900, loss[loss=0.2442, simple_loss=0.322, pruned_loss=0.08313, over 8467.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2928, pruned_loss=0.06492, over 1617552.31 frames. ], batch size: 25, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:31:26,323 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.692e+02 2.409e+02 2.783e+02 3.401e+02 8.568e+02, threshold=5.566e+02, percent-clipped=1.0 +2023-02-07 00:31:50,426 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:31:53,722 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 00:31:57,136 INFO [train.py:901] (1/4) Epoch 20, batch 2950, loss[loss=0.1666, simple_loss=0.25, pruned_loss=0.04162, over 5582.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2925, pruned_loss=0.06436, over 1616869.89 frames. ], batch size: 12, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:32:02,687 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6734, 1.4925, 4.8829, 1.7557, 4.3470, 4.0846, 4.4459, 4.2988], + device='cuda:1'), covar=tensor([0.0514, 0.4503, 0.0450, 0.3840, 0.1004, 0.0877, 0.0532, 0.0601], + device='cuda:1'), in_proj_covar=tensor([0.0607, 0.0627, 0.0675, 0.0609, 0.0691, 0.0597, 0.0593, 0.0659], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:32:08,262 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156542.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:32:11,758 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156547.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:32:17,784 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6983, 1.9332, 2.0078, 1.3257, 2.1761, 1.5201, 0.7090, 1.9156], + device='cuda:1'), covar=tensor([0.0518, 0.0334, 0.0276, 0.0551, 0.0390, 0.0755, 0.0774, 0.0280], + device='cuda:1'), in_proj_covar=tensor([0.0446, 0.0384, 0.0337, 0.0439, 0.0368, 0.0530, 0.0389, 0.0409], + device='cuda:1'), out_proj_covar=tensor([1.2015e-04, 1.0112e-04, 8.8875e-05, 1.1625e-04, 9.7303e-05, 1.5098e-04, + 1.0529e-04, 1.0906e-04], device='cuda:1') +2023-02-07 00:32:31,033 INFO [train.py:901] (1/4) Epoch 20, batch 3000, loss[loss=0.2324, simple_loss=0.307, pruned_loss=0.07892, over 8352.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.293, pruned_loss=0.06479, over 1617959.79 frames. ], batch size: 26, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:32:31,034 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 00:32:44,429 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3656, 1.4579, 1.2987, 1.7345, 0.8165, 1.1299, 1.2046, 1.3697], + device='cuda:1'), covar=tensor([0.0565, 0.0660, 0.0676, 0.0322, 0.0953, 0.1015, 0.0672, 0.0680], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0196, 0.0246, 0.0211, 0.0206, 0.0247, 0.0251, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 00:32:46,840 INFO [train.py:935] (1/4) Epoch 20, validation: loss=0.1756, simple_loss=0.2756, pruned_loss=0.03779, over 944034.00 frames. +2023-02-07 00:32:46,842 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 00:32:48,387 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=156578.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:32:51,796 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.571e+02 2.420e+02 3.007e+02 3.801e+02 6.408e+02, threshold=6.014e+02, percent-clipped=4.0 +2023-02-07 00:33:12,266 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8763, 1.6197, 1.7586, 1.4070, 0.9328, 1.5805, 1.6441, 1.6643], + device='cuda:1'), covar=tensor([0.0511, 0.1132, 0.1518, 0.1346, 0.0546, 0.1310, 0.0656, 0.0564], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0159, 0.0100, 0.0162, 0.0112, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 00:33:22,164 INFO [train.py:901] (1/4) Epoch 20, batch 3050, loss[loss=0.212, simple_loss=0.2972, pruned_loss=0.06346, over 8237.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.292, pruned_loss=0.06485, over 1614318.30 frames. ], batch size: 22, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:33:40,598 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=156652.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:33:57,465 INFO [train.py:901] (1/4) Epoch 20, batch 3100, loss[loss=0.2011, simple_loss=0.288, pruned_loss=0.05712, over 8291.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2907, pruned_loss=0.06414, over 1615237.37 frames. ], batch size: 23, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:34:02,301 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.428e+02 2.992e+02 3.732e+02 8.006e+02, threshold=5.985e+02, percent-clipped=5.0 +2023-02-07 00:34:02,449 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.4802, 1.1325, 3.7926, 1.5426, 3.0043, 2.9568, 3.3896, 3.3963], + device='cuda:1'), covar=tensor([0.1594, 0.7361, 0.1461, 0.5383, 0.2538, 0.2118, 0.1379, 0.1304], + device='cuda:1'), in_proj_covar=tensor([0.0613, 0.0632, 0.0681, 0.0614, 0.0694, 0.0600, 0.0598, 0.0662], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:34:09,239 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156693.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:34:10,727 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0761, 1.8525, 2.3816, 1.9904, 2.3344, 2.1706, 1.8987, 1.0737], + device='cuda:1'), covar=tensor([0.5246, 0.4428, 0.1786, 0.3427, 0.2324, 0.2859, 0.1840, 0.4982], + device='cuda:1'), in_proj_covar=tensor([0.0933, 0.0963, 0.0790, 0.0929, 0.0986, 0.0881, 0.0740, 0.0815], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 00:34:12,936 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 +2023-02-07 00:34:28,114 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5352, 2.5559, 1.8686, 2.2407, 2.2352, 1.6077, 2.0467, 2.1458], + device='cuda:1'), covar=tensor([0.1457, 0.0439, 0.1202, 0.0622, 0.0693, 0.1544, 0.0961, 0.0920], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0237, 0.0331, 0.0307, 0.0299, 0.0336, 0.0344, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 00:34:31,976 INFO [train.py:901] (1/4) Epoch 20, batch 3150, loss[loss=0.18, simple_loss=0.2616, pruned_loss=0.04918, over 7640.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2906, pruned_loss=0.0642, over 1615870.11 frames. ], batch size: 19, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:34:36,352 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 00:34:58,561 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0136, 1.6964, 1.8176, 1.4716, 0.9557, 1.6194, 1.7603, 1.7141], + device='cuda:1'), covar=tensor([0.0501, 0.1167, 0.1585, 0.1339, 0.0587, 0.1360, 0.0650, 0.0595], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0158, 0.0100, 0.0161, 0.0112, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 00:35:01,275 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156767.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:35:01,312 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156767.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:35:07,246 INFO [train.py:901] (1/4) Epoch 20, batch 3200, loss[loss=0.2161, simple_loss=0.2867, pruned_loss=0.07274, over 7184.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2913, pruned_loss=0.06464, over 1612608.92 frames. ], batch size: 16, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:35:11,882 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.830e+02 2.338e+02 2.875e+02 3.612e+02 1.133e+03, threshold=5.749e+02, percent-clipped=4.0 +2023-02-07 00:35:21,727 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5592, 1.5410, 1.8448, 1.2890, 1.2658, 1.7981, 0.1840, 1.2650], + device='cuda:1'), covar=tensor([0.1796, 0.1458, 0.0419, 0.1071, 0.2622, 0.0511, 0.2047, 0.1260], + device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0194, 0.0124, 0.0220, 0.0268, 0.0133, 0.0168, 0.0188], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 00:35:26,512 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156803.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:35:41,820 INFO [train.py:901] (1/4) Epoch 20, batch 3250, loss[loss=0.1765, simple_loss=0.2534, pruned_loss=0.04975, over 7435.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2908, pruned_loss=0.06426, over 1611600.02 frames. ], batch size: 17, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:35:43,292 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156828.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:36:17,055 INFO [train.py:901] (1/4) Epoch 20, batch 3300, loss[loss=0.2031, simple_loss=0.2957, pruned_loss=0.05524, over 8251.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2915, pruned_loss=0.0646, over 1615628.39 frames. ], batch size: 24, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:36:21,773 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.341e+02 2.967e+02 3.887e+02 7.432e+02, threshold=5.934e+02, percent-clipped=7.0 +2023-02-07 00:36:35,544 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156903.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:36:51,531 INFO [train.py:901] (1/4) Epoch 20, batch 3350, loss[loss=0.1789, simple_loss=0.2605, pruned_loss=0.0487, over 7414.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2911, pruned_loss=0.06442, over 1612267.88 frames. ], batch size: 17, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:37:07,172 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156949.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:37:25,860 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156974.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:37:27,106 INFO [train.py:901] (1/4) Epoch 20, batch 3400, loss[loss=0.2166, simple_loss=0.3104, pruned_loss=0.06141, over 8355.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2907, pruned_loss=0.06425, over 1615533.14 frames. ], batch size: 26, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:37:31,900 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.508e+02 3.011e+02 3.882e+02 8.239e+02, threshold=6.022e+02, percent-clipped=6.0 +2023-02-07 00:38:01,311 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157023.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:38:03,928 INFO [train.py:901] (1/4) Epoch 20, batch 3450, loss[loss=0.2154, simple_loss=0.2866, pruned_loss=0.07205, over 8233.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2908, pruned_loss=0.06449, over 1615766.41 frames. ], batch size: 22, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:38:05,360 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157028.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:38:18,896 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157048.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:38:38,197 INFO [train.py:901] (1/4) Epoch 20, batch 3500, loss[loss=0.1998, simple_loss=0.2976, pruned_loss=0.05097, over 8101.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2919, pruned_loss=0.06527, over 1617537.81 frames. ], batch size: 23, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:38:43,591 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.548e+02 3.004e+02 3.939e+02 7.448e+02, threshold=6.007e+02, percent-clipped=9.0 +2023-02-07 00:39:02,221 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 00:39:03,067 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157111.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:39:13,060 INFO [train.py:901] (1/4) Epoch 20, batch 3550, loss[loss=0.2251, simple_loss=0.2969, pruned_loss=0.07662, over 8080.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2918, pruned_loss=0.06528, over 1620563.48 frames. ], batch size: 21, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:39:37,685 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157160.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:39:43,854 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1127, 1.3652, 4.3144, 1.5688, 3.7531, 3.6327, 3.9235, 3.7853], + device='cuda:1'), covar=tensor([0.0714, 0.4832, 0.0550, 0.4022, 0.1225, 0.0958, 0.0631, 0.0724], + device='cuda:1'), in_proj_covar=tensor([0.0608, 0.0631, 0.0680, 0.0610, 0.0692, 0.0598, 0.0595, 0.0659], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:39:48,307 INFO [train.py:901] (1/4) Epoch 20, batch 3600, loss[loss=0.2527, simple_loss=0.3297, pruned_loss=0.08787, over 8352.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2911, pruned_loss=0.06459, over 1620742.66 frames. ], batch size: 24, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:39:49,146 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9148, 1.4113, 1.7026, 1.3012, 0.9557, 1.4360, 1.7101, 1.3555], + device='cuda:1'), covar=tensor([0.0514, 0.1296, 0.1637, 0.1478, 0.0591, 0.1515, 0.0687, 0.0689], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0152, 0.0190, 0.0158, 0.0100, 0.0161, 0.0112, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 00:39:53,036 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.446e+02 2.923e+02 3.668e+02 9.434e+02, threshold=5.847e+02, percent-clipped=4.0 +2023-02-07 00:40:24,187 INFO [train.py:901] (1/4) Epoch 20, batch 3650, loss[loss=0.2264, simple_loss=0.2948, pruned_loss=0.07898, over 7276.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2899, pruned_loss=0.06401, over 1615332.59 frames. ], batch size: 16, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:40:24,361 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157226.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:40:38,564 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157247.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:40:44,448 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1589, 3.8888, 2.4738, 2.7467, 3.2053, 2.0499, 3.2843, 3.1509], + device='cuda:1'), covar=tensor([0.1491, 0.0314, 0.0992, 0.0756, 0.0575, 0.1377, 0.0846, 0.0934], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0235, 0.0330, 0.0307, 0.0299, 0.0335, 0.0343, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 00:40:58,619 INFO [train.py:901] (1/4) Epoch 20, batch 3700, loss[loss=0.2152, simple_loss=0.2897, pruned_loss=0.07034, over 8339.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2895, pruned_loss=0.06358, over 1615372.12 frames. ], batch size: 26, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:41:03,187 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.545e+02 3.038e+02 3.849e+02 9.039e+02, threshold=6.076e+02, percent-clipped=6.0 +2023-02-07 00:41:05,243 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 00:41:33,573 INFO [train.py:901] (1/4) Epoch 20, batch 3750, loss[loss=0.191, simple_loss=0.2619, pruned_loss=0.06004, over 7804.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2895, pruned_loss=0.06412, over 1609337.44 frames. ], batch size: 19, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:41:58,945 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157362.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:42:05,451 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157372.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:42:06,810 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157374.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:42:07,964 INFO [train.py:901] (1/4) Epoch 20, batch 3800, loss[loss=0.2202, simple_loss=0.2987, pruned_loss=0.07085, over 8327.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2888, pruned_loss=0.06353, over 1610865.29 frames. ], batch size: 25, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:42:11,380 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3684, 1.6021, 4.5683, 1.8323, 4.0211, 3.8221, 4.1827, 4.0444], + device='cuda:1'), covar=tensor([0.0552, 0.4310, 0.0483, 0.3986, 0.1031, 0.0922, 0.0517, 0.0609], + device='cuda:1'), in_proj_covar=tensor([0.0612, 0.0634, 0.0684, 0.0614, 0.0694, 0.0604, 0.0598, 0.0663], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:42:12,512 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.302e+02 2.981e+02 3.884e+02 7.104e+02, threshold=5.962e+02, percent-clipped=4.0 +2023-02-07 00:42:21,034 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6056, 1.9380, 2.1569, 1.0956, 2.2512, 1.5013, 0.6398, 1.8891], + device='cuda:1'), covar=tensor([0.0619, 0.0366, 0.0271, 0.0664, 0.0349, 0.0894, 0.0884, 0.0295], + device='cuda:1'), in_proj_covar=tensor([0.0446, 0.0388, 0.0339, 0.0441, 0.0372, 0.0533, 0.0393, 0.0413], + device='cuda:1'), out_proj_covar=tensor([1.2026e-04, 1.0208e-04, 8.9457e-05, 1.1672e-04, 9.8420e-05, 1.5163e-04, + 1.0626e-04, 1.1019e-04], device='cuda:1') +2023-02-07 00:42:25,029 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157400.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:42:39,605 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157422.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:42:42,777 INFO [train.py:901] (1/4) Epoch 20, batch 3850, loss[loss=0.1794, simple_loss=0.2693, pruned_loss=0.04479, over 8492.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2902, pruned_loss=0.06449, over 1612486.51 frames. ], batch size: 28, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:43:09,749 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 00:43:17,669 INFO [train.py:901] (1/4) Epoch 20, batch 3900, loss[loss=0.235, simple_loss=0.3132, pruned_loss=0.07836, over 8439.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2908, pruned_loss=0.06444, over 1614560.08 frames. ], batch size: 27, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:43:21,793 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157482.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:43:22,211 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.513e+02 3.153e+02 3.900e+02 7.255e+02, threshold=6.305e+02, percent-clipped=5.0 +2023-02-07 00:43:24,970 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157487.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:43:37,149 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157504.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:43:39,431 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157507.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:43:51,282 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157524.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:43:52,542 INFO [train.py:901] (1/4) Epoch 20, batch 3950, loss[loss=0.2301, simple_loss=0.3081, pruned_loss=0.07607, over 8681.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2903, pruned_loss=0.06417, over 1612184.82 frames. ], batch size: 34, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:44:21,478 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4314, 2.6518, 3.0646, 1.4960, 3.2535, 1.9918, 1.4624, 2.4044], + device='cuda:1'), covar=tensor([0.0685, 0.0346, 0.0291, 0.0803, 0.0451, 0.0838, 0.0984, 0.0484], + device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0390, 0.0341, 0.0444, 0.0373, 0.0535, 0.0395, 0.0414], + device='cuda:1'), out_proj_covar=tensor([1.2124e-04, 1.0247e-04, 9.0116e-05, 1.1761e-04, 9.8803e-05, 1.5224e-04, + 1.0693e-04, 1.1042e-04], device='cuda:1') +2023-02-07 00:44:28,450 INFO [train.py:901] (1/4) Epoch 20, batch 4000, loss[loss=0.2112, simple_loss=0.2882, pruned_loss=0.06715, over 8465.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2895, pruned_loss=0.06298, over 1613816.06 frames. ], batch size: 27, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:44:33,892 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 2.441e+02 3.259e+02 3.960e+02 7.383e+02, threshold=6.518e+02, percent-clipped=3.0 +2023-02-07 00:44:34,867 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6479, 2.4639, 3.3928, 2.6263, 3.0637, 2.5884, 2.4038, 1.9627], + device='cuda:1'), covar=tensor([0.5087, 0.4821, 0.1797, 0.3581, 0.2519, 0.2835, 0.1806, 0.5193], + device='cuda:1'), in_proj_covar=tensor([0.0929, 0.0959, 0.0784, 0.0924, 0.0980, 0.0878, 0.0733, 0.0811], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 00:44:36,131 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:44:57,800 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157618.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:44:58,428 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157619.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:45:03,491 INFO [train.py:901] (1/4) Epoch 20, batch 4050, loss[loss=0.2377, simple_loss=0.3178, pruned_loss=0.07878, over 8250.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2905, pruned_loss=0.06388, over 1612420.66 frames. ], batch size: 22, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:45:12,623 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 00:45:15,104 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:45:26,959 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7730, 1.8760, 1.6887, 2.2948, 0.9791, 1.4996, 1.6682, 1.8844], + device='cuda:1'), covar=tensor([0.0783, 0.0807, 0.0936, 0.0443, 0.1322, 0.1393, 0.0876, 0.0754], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0199, 0.0249, 0.0214, 0.0207, 0.0251, 0.0255, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 00:45:33,631 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4238, 1.4874, 1.4497, 1.8290, 0.7369, 1.3108, 1.2875, 1.4957], + device='cuda:1'), covar=tensor([0.0881, 0.0819, 0.0969, 0.0511, 0.1238, 0.1431, 0.0822, 0.0720], + device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0199, 0.0249, 0.0214, 0.0207, 0.0251, 0.0255, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 00:45:38,035 INFO [train.py:901] (1/4) Epoch 20, batch 4100, loss[loss=0.23, simple_loss=0.3206, pruned_loss=0.06972, over 8038.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2919, pruned_loss=0.06506, over 1608903.95 frames. ], batch size: 22, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:45:42,595 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.673e+02 2.468e+02 3.178e+02 4.268e+02 8.149e+02, threshold=6.355e+02, percent-clipped=4.0 +2023-02-07 00:45:52,658 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.85 vs. limit=5.0 +2023-02-07 00:46:07,480 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157718.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:46:12,789 INFO [train.py:901] (1/4) Epoch 20, batch 4150, loss[loss=0.245, simple_loss=0.3139, pruned_loss=0.08803, over 6796.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2924, pruned_loss=0.06508, over 1610490.62 frames. ], batch size: 71, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:46:13,016 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7361, 1.6918, 2.3741, 1.6340, 1.2917, 2.3354, 0.5108, 1.3910], + device='cuda:1'), covar=tensor([0.1851, 0.1498, 0.0334, 0.1258, 0.3133, 0.0448, 0.2547, 0.1610], + device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0195, 0.0124, 0.0219, 0.0268, 0.0133, 0.0168, 0.0189], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 00:46:25,110 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157743.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:46:25,656 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157744.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:46:40,584 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157766.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:46:41,942 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:46:46,942 INFO [train.py:901] (1/4) Epoch 20, batch 4200, loss[loss=0.2533, simple_loss=0.3243, pruned_loss=0.09119, over 8460.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2918, pruned_loss=0.06485, over 1610613.40 frames. ], batch size: 25, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:46:52,351 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.385e+02 2.811e+02 3.577e+02 7.269e+02, threshold=5.621e+02, percent-clipped=2.0 +2023-02-07 00:47:08,781 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 00:47:23,328 INFO [train.py:901] (1/4) Epoch 20, batch 4250, loss[loss=0.2049, simple_loss=0.2827, pruned_loss=0.06353, over 8240.00 frames. ], tot_loss[loss=0.21, simple_loss=0.291, pruned_loss=0.06445, over 1611498.40 frames. ], batch size: 22, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:47:28,295 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157833.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:47:32,309 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 00:47:46,378 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157859.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:47:53,368 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157868.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:47:55,490 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157871.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:47:58,150 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157875.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:47:58,653 INFO [train.py:901] (1/4) Epoch 20, batch 4300, loss[loss=0.202, simple_loss=0.2956, pruned_loss=0.05426, over 8336.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2909, pruned_loss=0.06402, over 1615827.78 frames. ], batch size: 26, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:48:02,061 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157881.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:48:03,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.270e+02 2.745e+02 3.400e+02 8.203e+02, threshold=5.491e+02, percent-clipped=7.0 +2023-02-07 00:48:15,330 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157900.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:48:20,878 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 00:48:33,490 INFO [train.py:901] (1/4) Epoch 20, batch 4350, loss[loss=0.17, simple_loss=0.2513, pruned_loss=0.0444, over 7794.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.29, pruned_loss=0.06389, over 1614737.52 frames. ], batch size: 19, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:48:36,276 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157930.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:49:00,924 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 00:49:04,113 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 00:49:08,270 INFO [train.py:901] (1/4) Epoch 20, batch 4400, loss[loss=0.1859, simple_loss=0.2641, pruned_loss=0.05389, over 7434.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2887, pruned_loss=0.06312, over 1611183.95 frames. ], batch size: 17, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:49:13,805 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.443e+02 2.894e+02 3.714e+02 1.238e+03, threshold=5.788e+02, percent-clipped=6.0 +2023-02-07 00:49:14,009 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157983.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:49:44,326 INFO [train.py:901] (1/4) Epoch 20, batch 4450, loss[loss=0.2078, simple_loss=0.281, pruned_loss=0.0673, over 7971.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2895, pruned_loss=0.06365, over 1608290.18 frames. ], batch size: 21, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:49:45,694 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 00:49:57,302 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158045.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:50:17,226 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-07 00:50:18,896 INFO [train.py:901] (1/4) Epoch 20, batch 4500, loss[loss=0.1982, simple_loss=0.2749, pruned_loss=0.06073, over 7926.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2901, pruned_loss=0.06399, over 1609451.65 frames. ], batch size: 20, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:50:23,587 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.382e+02 2.908e+02 3.384e+02 7.082e+02, threshold=5.816e+02, percent-clipped=5.0 +2023-02-07 00:50:27,959 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158089.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:50:35,485 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1876, 1.3320, 4.3446, 1.7223, 3.9195, 3.6326, 3.9530, 3.8497], + device='cuda:1'), covar=tensor([0.0514, 0.4540, 0.0552, 0.3708, 0.1049, 0.0915, 0.0519, 0.0609], + device='cuda:1'), in_proj_covar=tensor([0.0611, 0.0633, 0.0684, 0.0615, 0.0695, 0.0603, 0.0597, 0.0663], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:50:39,142 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 00:50:45,177 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158114.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:50:45,857 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158115.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:50:53,562 INFO [train.py:901] (1/4) Epoch 20, batch 4550, loss[loss=0.1798, simple_loss=0.2656, pruned_loss=0.04701, over 7921.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2907, pruned_loss=0.06438, over 1610195.06 frames. ], batch size: 20, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:51:01,044 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158137.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:51:02,890 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158140.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:51:02,944 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158140.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:51:18,404 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158162.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:51:28,240 INFO [train.py:901] (1/4) Epoch 20, batch 4600, loss[loss=0.2011, simple_loss=0.2917, pruned_loss=0.05523, over 8196.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2908, pruned_loss=0.06433, over 1611511.14 frames. ], batch size: 23, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:51:29,052 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7892, 1.3572, 3.9754, 1.4843, 3.5155, 3.3151, 3.5998, 3.4842], + device='cuda:1'), covar=tensor([0.0759, 0.4418, 0.0649, 0.4024, 0.1242, 0.0999, 0.0670, 0.0767], + device='cuda:1'), in_proj_covar=tensor([0.0610, 0.0631, 0.0681, 0.0614, 0.0693, 0.0602, 0.0596, 0.0660], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:51:32,830 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.506e+02 3.217e+02 3.763e+02 8.986e+02, threshold=6.435e+02, percent-clipped=3.0 +2023-02-07 00:51:54,921 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=158215.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:52:03,091 INFO [train.py:901] (1/4) Epoch 20, batch 4650, loss[loss=0.2052, simple_loss=0.2859, pruned_loss=0.0622, over 8456.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2901, pruned_loss=0.0643, over 1610755.93 frames. ], batch size: 49, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:52:12,074 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158239.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:52:29,426 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3861, 1.4058, 4.5703, 1.6911, 4.0492, 3.8484, 4.1788, 4.0355], + device='cuda:1'), covar=tensor([0.0595, 0.4422, 0.0471, 0.3820, 0.1059, 0.0874, 0.0518, 0.0646], + device='cuda:1'), in_proj_covar=tensor([0.0612, 0.0634, 0.0685, 0.0616, 0.0695, 0.0605, 0.0598, 0.0663], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:52:30,102 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158264.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:52:37,869 INFO [train.py:901] (1/4) Epoch 20, batch 4700, loss[loss=0.273, simple_loss=0.3423, pruned_loss=0.1018, over 6651.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2899, pruned_loss=0.06405, over 1610658.41 frames. ], batch size: 71, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:52:42,600 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.408e+02 3.012e+02 4.119e+02 1.091e+03, threshold=6.025e+02, percent-clipped=3.0 +2023-02-07 00:52:53,700 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2338, 3.1142, 2.9270, 1.6095, 2.8559, 2.9222, 2.8475, 2.8069], + device='cuda:1'), covar=tensor([0.1109, 0.0811, 0.1291, 0.4581, 0.1164, 0.1414, 0.1565, 0.1145], + device='cuda:1'), in_proj_covar=tensor([0.0519, 0.0422, 0.0429, 0.0528, 0.0418, 0.0430, 0.0418, 0.0373], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:52:55,084 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9210, 1.3352, 3.1740, 1.2254, 2.4710, 2.4923, 2.9078, 2.8907], + device='cuda:1'), covar=tensor([0.1689, 0.5997, 0.1871, 0.5356, 0.3020, 0.2444, 0.1421, 0.1473], + device='cuda:1'), in_proj_covar=tensor([0.0612, 0.0635, 0.0687, 0.0615, 0.0697, 0.0604, 0.0599, 0.0664], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:52:55,798 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158301.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:53:12,689 INFO [train.py:901] (1/4) Epoch 20, batch 4750, loss[loss=0.2472, simple_loss=0.324, pruned_loss=0.08513, over 8330.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2888, pruned_loss=0.06366, over 1610415.59 frames. ], batch size: 26, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:53:12,918 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158326.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:53:15,561 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158330.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:53:21,788 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4189, 2.3436, 1.5441, 2.2488, 2.0032, 1.2388, 1.9330, 2.1629], + device='cuda:1'), covar=tensor([0.1831, 0.0587, 0.1596, 0.0731, 0.1031, 0.2238, 0.1342, 0.1084], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0236, 0.0332, 0.0309, 0.0303, 0.0338, 0.0343, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 00:53:24,190 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-07 00:53:40,946 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 00:53:43,681 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 00:53:48,277 INFO [train.py:901] (1/4) Epoch 20, batch 4800, loss[loss=0.2263, simple_loss=0.3023, pruned_loss=0.07514, over 8495.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2899, pruned_loss=0.06391, over 1610914.26 frames. ], batch size: 39, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:53:52,906 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.386e+02 2.729e+02 3.445e+02 7.258e+02, threshold=5.458e+02, percent-clipped=2.0 +2023-02-07 00:54:06,942 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158402.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:54:22,900 INFO [train.py:901] (1/4) Epoch 20, batch 4850, loss[loss=0.3195, simple_loss=0.3678, pruned_loss=0.1356, over 6664.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2907, pruned_loss=0.0647, over 1613231.07 frames. ], batch size: 71, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:54:33,551 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 00:54:57,241 INFO [train.py:901] (1/4) Epoch 20, batch 4900, loss[loss=0.2044, simple_loss=0.311, pruned_loss=0.04896, over 8191.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2913, pruned_loss=0.06444, over 1612816.19 frames. ], batch size: 23, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:55:02,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.481e+02 3.123e+02 4.208e+02 8.958e+02, threshold=6.246e+02, percent-clipped=7.0 +2023-02-07 00:55:03,229 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=158484.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:55:28,957 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4666, 1.5848, 1.4986, 1.8437, 0.7714, 1.3544, 1.3907, 1.5782], + device='cuda:1'), covar=tensor([0.0759, 0.0719, 0.0916, 0.0479, 0.1067, 0.1303, 0.0692, 0.0652], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0196, 0.0246, 0.0213, 0.0204, 0.0248, 0.0251, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 00:55:32,873 INFO [train.py:901] (1/4) Epoch 20, batch 4950, loss[loss=0.2855, simple_loss=0.3632, pruned_loss=0.1039, over 8671.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2916, pruned_loss=0.06469, over 1609431.33 frames. ], batch size: 34, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:56:07,761 INFO [train.py:901] (1/4) Epoch 20, batch 5000, loss[loss=0.2243, simple_loss=0.306, pruned_loss=0.07136, over 8367.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2922, pruned_loss=0.06478, over 1614974.90 frames. ], batch size: 24, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:56:12,218 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.361e+02 2.881e+02 3.667e+02 7.563e+02, threshold=5.761e+02, percent-clipped=2.0 +2023-02-07 00:56:14,509 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:56:23,811 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158599.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:56:32,786 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158611.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:56:42,905 INFO [train.py:901] (1/4) Epoch 20, batch 5050, loss[loss=0.1987, simple_loss=0.2905, pruned_loss=0.05345, over 8325.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.292, pruned_loss=0.06482, over 1612578.88 frames. ], batch size: 25, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:57:10,200 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 00:57:17,769 INFO [train.py:901] (1/4) Epoch 20, batch 5100, loss[loss=0.2115, simple_loss=0.302, pruned_loss=0.06049, over 8094.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2927, pruned_loss=0.06492, over 1616471.64 frames. ], batch size: 21, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:57:23,332 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.670e+02 3.233e+02 3.910e+02 8.185e+02, threshold=6.466e+02, percent-clipped=7.0 +2023-02-07 00:57:38,618 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 00:57:53,857 INFO [train.py:901] (1/4) Epoch 20, batch 5150, loss[loss=0.1883, simple_loss=0.2554, pruned_loss=0.06065, over 7169.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2932, pruned_loss=0.06515, over 1620073.68 frames. ], batch size: 16, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:57:55,011 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 00:58:08,320 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=158746.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:58:28,394 INFO [train.py:901] (1/4) Epoch 20, batch 5200, loss[loss=0.1894, simple_loss=0.2763, pruned_loss=0.05123, over 7963.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2924, pruned_loss=0.06475, over 1613818.44 frames. ], batch size: 21, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:58:30,739 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0753, 1.2425, 1.1801, 0.6880, 1.2095, 1.0260, 0.0665, 1.2098], + device='cuda:1'), covar=tensor([0.0359, 0.0327, 0.0290, 0.0507, 0.0361, 0.0805, 0.0740, 0.0276], + device='cuda:1'), in_proj_covar=tensor([0.0454, 0.0389, 0.0340, 0.0442, 0.0372, 0.0535, 0.0395, 0.0416], + device='cuda:1'), out_proj_covar=tensor([1.2222e-04, 1.0224e-04, 8.9611e-05, 1.1686e-04, 9.8562e-05, 1.5211e-04, + 1.0678e-04, 1.1088e-04], device='cuda:1') +2023-02-07 00:58:31,274 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2248, 3.1334, 2.9067, 1.4823, 2.9045, 2.8715, 2.8901, 2.7940], + device='cuda:1'), covar=tensor([0.1153, 0.0789, 0.1333, 0.4592, 0.1162, 0.1207, 0.1567, 0.1015], + device='cuda:1'), in_proj_covar=tensor([0.0521, 0.0425, 0.0436, 0.0533, 0.0422, 0.0435, 0.0422, 0.0377], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 00:58:33,212 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.433e+02 2.837e+02 3.461e+02 7.505e+02, threshold=5.673e+02, percent-clipped=2.0 +2023-02-07 00:58:41,616 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158795.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:58:46,635 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158801.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:59:03,969 INFO [train.py:901] (1/4) Epoch 20, batch 5250, loss[loss=0.2261, simple_loss=0.3023, pruned_loss=0.07498, over 8246.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2918, pruned_loss=0.06423, over 1615727.86 frames. ], batch size: 24, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:59:06,915 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3291, 1.2961, 2.3801, 1.2533, 2.2025, 2.5132, 2.7115, 2.1202], + device='cuda:1'), covar=tensor([0.1203, 0.1443, 0.0435, 0.2060, 0.0673, 0.0400, 0.0573, 0.0683], + device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0318, 0.0285, 0.0312, 0.0301, 0.0261, 0.0407, 0.0302], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 00:59:11,275 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 00:59:22,844 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158853.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:59:24,328 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158855.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:59:28,241 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158861.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:59:38,589 INFO [train.py:901] (1/4) Epoch 20, batch 5300, loss[loss=0.2399, simple_loss=0.2993, pruned_loss=0.09029, over 7818.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2906, pruned_loss=0.06371, over 1614225.80 frames. ], batch size: 20, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 00:59:41,427 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158880.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:59:43,360 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.336e+02 2.792e+02 3.296e+02 7.091e+02, threshold=5.585e+02, percent-clipped=2.0 +2023-02-07 01:00:13,210 INFO [train.py:901] (1/4) Epoch 20, batch 5350, loss[loss=0.2248, simple_loss=0.3101, pruned_loss=0.06974, over 8463.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2902, pruned_loss=0.06382, over 1606863.21 frames. ], batch size: 29, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:00:22,118 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 01:00:48,005 INFO [train.py:901] (1/4) Epoch 20, batch 5400, loss[loss=0.2182, simple_loss=0.3054, pruned_loss=0.06556, over 8462.00 frames. ], tot_loss[loss=0.21, simple_loss=0.291, pruned_loss=0.06448, over 1607533.23 frames. ], batch size: 25, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:00:52,647 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.394e+02 2.966e+02 3.887e+02 6.953e+02, threshold=5.932e+02, percent-clipped=4.0 +2023-02-07 01:01:22,913 INFO [train.py:901] (1/4) Epoch 20, batch 5450, loss[loss=0.1899, simple_loss=0.2791, pruned_loss=0.05036, over 8247.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2907, pruned_loss=0.06437, over 1611191.94 frames. ], batch size: 24, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:01:23,771 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2362, 1.3211, 3.3799, 1.0870, 3.0170, 2.8508, 3.0974, 2.9826], + device='cuda:1'), covar=tensor([0.0789, 0.3818, 0.0756, 0.3800, 0.1286, 0.1044, 0.0743, 0.0885], + device='cuda:1'), in_proj_covar=tensor([0.0610, 0.0630, 0.0677, 0.0612, 0.0691, 0.0596, 0.0596, 0.0664], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:01:57,430 INFO [train.py:901] (1/4) Epoch 20, batch 5500, loss[loss=0.2051, simple_loss=0.2927, pruned_loss=0.05877, over 8249.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2904, pruned_loss=0.06387, over 1611456.07 frames. ], batch size: 22, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:02:00,083 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 01:02:02,834 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.672e+02 2.298e+02 2.656e+02 3.222e+02 6.486e+02, threshold=5.312e+02, percent-clipped=1.0 +2023-02-07 01:02:05,771 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159087.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:10,089 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3966, 1.5655, 2.1964, 1.3366, 1.5769, 1.6542, 1.4723, 1.6278], + device='cuda:1'), covar=tensor([0.2054, 0.2584, 0.0908, 0.4517, 0.1964, 0.3456, 0.2427, 0.1983], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0590, 0.0554, 0.0633, 0.0643, 0.0589, 0.0528, 0.0629], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:02:19,886 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0781, 1.8546, 2.4010, 2.0347, 2.3246, 2.0804, 1.9027, 1.5963], + device='cuda:1'), covar=tensor([0.3775, 0.3935, 0.1506, 0.2629, 0.1769, 0.2377, 0.1565, 0.3769], + device='cuda:1'), in_proj_covar=tensor([0.0929, 0.0962, 0.0787, 0.0927, 0.0980, 0.0877, 0.0734, 0.0811], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 01:02:27,309 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159117.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:32,435 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1852, 1.6710, 4.3806, 1.6437, 3.9488, 3.7287, 3.9897, 3.9065], + device='cuda:1'), covar=tensor([0.0617, 0.4078, 0.0539, 0.4235, 0.1114, 0.0895, 0.0604, 0.0657], + device='cuda:1'), in_proj_covar=tensor([0.0609, 0.0628, 0.0674, 0.0610, 0.0690, 0.0595, 0.0594, 0.0662], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:02:32,992 INFO [train.py:901] (1/4) Epoch 20, batch 5550, loss[loss=0.1792, simple_loss=0.2582, pruned_loss=0.05012, over 7724.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2903, pruned_loss=0.06358, over 1610082.28 frames. ], batch size: 18, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:02:41,928 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:44,149 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159142.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:46,023 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159145.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:59,834 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7687, 1.7028, 1.9304, 1.5923, 1.1120, 1.7851, 2.3806, 2.0021], + device='cuda:1'), covar=tensor([0.0502, 0.1222, 0.1590, 0.1396, 0.0624, 0.1394, 0.0602, 0.0614], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0099, 0.0161, 0.0112, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 01:03:01,231 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0539, 2.3822, 3.6950, 1.7707, 1.6435, 3.6879, 0.8740, 2.0553], + device='cuda:1'), covar=tensor([0.1417, 0.1125, 0.0194, 0.1833, 0.2848, 0.0253, 0.1952, 0.1291], + device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0195, 0.0125, 0.0219, 0.0269, 0.0133, 0.0167, 0.0189], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 01:03:08,171 INFO [train.py:901] (1/4) Epoch 20, batch 5600, loss[loss=0.2355, simple_loss=0.2979, pruned_loss=0.08658, over 7207.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2898, pruned_loss=0.06356, over 1608065.63 frames. ], batch size: 16, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:03:09,020 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159177.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:03:12,919 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.419e+02 2.780e+02 3.445e+02 7.739e+02, threshold=5.561e+02, percent-clipped=2.0 +2023-02-07 01:03:15,215 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1849, 1.5111, 4.4563, 1.9549, 2.4758, 5.1095, 5.0706, 4.3724], + device='cuda:1'), covar=tensor([0.1201, 0.1846, 0.0289, 0.1964, 0.1118, 0.0163, 0.0357, 0.0541], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0320, 0.0288, 0.0315, 0.0305, 0.0263, 0.0412, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 01:03:23,294 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159197.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 01:03:30,905 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8569, 1.7001, 2.4810, 1.4001, 1.1729, 2.4316, 0.3780, 1.3319], + device='cuda:1'), covar=tensor([0.1888, 0.1379, 0.0321, 0.1457, 0.3100, 0.0331, 0.2396, 0.1551], + device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0196, 0.0125, 0.0219, 0.0270, 0.0134, 0.0168, 0.0190], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 01:03:43,998 INFO [train.py:901] (1/4) Epoch 20, batch 5650, loss[loss=0.2199, simple_loss=0.2981, pruned_loss=0.07083, over 7807.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2908, pruned_loss=0.06409, over 1609258.87 frames. ], batch size: 19, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:04:03,415 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159254.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:04:04,635 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 01:04:07,512 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159260.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:04:19,061 INFO [train.py:901] (1/4) Epoch 20, batch 5700, loss[loss=0.184, simple_loss=0.2784, pruned_loss=0.04482, over 8495.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2915, pruned_loss=0.06414, over 1609371.34 frames. ], batch size: 26, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:04:25,334 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.576e+02 3.260e+02 4.013e+02 6.441e+02, threshold=6.520e+02, percent-clipped=4.0 +2023-02-07 01:04:42,294 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159308.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:04:45,035 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159312.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 01:04:50,568 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0325, 1.3967, 3.4670, 1.6499, 2.2533, 3.8766, 3.9501, 3.3160], + device='cuda:1'), covar=tensor([0.1167, 0.1928, 0.0390, 0.1982, 0.1238, 0.0215, 0.0508, 0.0539], + device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0318, 0.0286, 0.0313, 0.0304, 0.0261, 0.0410, 0.0302], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 01:04:54,518 INFO [train.py:901] (1/4) Epoch 20, batch 5750, loss[loss=0.1828, simple_loss=0.2593, pruned_loss=0.05319, over 7656.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.291, pruned_loss=0.06382, over 1610554.99 frames. ], batch size: 19, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:05:09,319 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 01:05:21,858 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9943, 1.5860, 3.4497, 1.5522, 2.3307, 3.8027, 3.9197, 3.3077], + device='cuda:1'), covar=tensor([0.1083, 0.1646, 0.0342, 0.1988, 0.1008, 0.0214, 0.0499, 0.0481], + device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0316, 0.0284, 0.0311, 0.0302, 0.0260, 0.0407, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 01:05:29,354 INFO [train.py:901] (1/4) Epoch 20, batch 5800, loss[loss=0.2176, simple_loss=0.2974, pruned_loss=0.06894, over 8459.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2898, pruned_loss=0.06364, over 1607160.64 frames. ], batch size: 27, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:05:35,563 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.438e+02 2.992e+02 3.849e+02 1.447e+03, threshold=5.984e+02, percent-clipped=4.0 +2023-02-07 01:06:04,883 INFO [train.py:901] (1/4) Epoch 20, batch 5850, loss[loss=0.1751, simple_loss=0.2624, pruned_loss=0.0439, over 7961.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2901, pruned_loss=0.06372, over 1608992.53 frames. ], batch size: 21, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:06:08,465 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159431.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:06:09,313 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5088, 1.9013, 3.0499, 1.3893, 2.2024, 1.9978, 1.6400, 2.2314], + device='cuda:1'), covar=tensor([0.2073, 0.2661, 0.0902, 0.4766, 0.2074, 0.3269, 0.2389, 0.2448], + device='cuda:1'), in_proj_covar=tensor([0.0522, 0.0593, 0.0556, 0.0636, 0.0647, 0.0592, 0.0531, 0.0632], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:06:39,993 INFO [train.py:901] (1/4) Epoch 20, batch 5900, loss[loss=0.1916, simple_loss=0.2699, pruned_loss=0.05669, over 8249.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2891, pruned_loss=0.06316, over 1608360.61 frames. ], batch size: 22, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:06:45,626 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.454e+02 2.951e+02 3.822e+02 7.063e+02, threshold=5.901e+02, percent-clipped=2.0 +2023-02-07 01:07:04,121 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159510.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:08,830 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159516.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:12,108 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159521.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:15,452 INFO [train.py:901] (1/4) Epoch 20, batch 5950, loss[loss=0.1939, simple_loss=0.2861, pruned_loss=0.05092, over 8107.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2903, pruned_loss=0.06374, over 1608245.10 frames. ], batch size: 23, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:07:21,779 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159535.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:26,380 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159541.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:29,667 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159546.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:45,569 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159568.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:07:50,936 INFO [train.py:901] (1/4) Epoch 20, batch 6000, loss[loss=0.1823, simple_loss=0.2736, pruned_loss=0.04547, over 8509.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2901, pruned_loss=0.06339, over 1612212.27 frames. ], batch size: 26, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:07:50,937 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 01:08:01,340 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1314, 1.8333, 2.3860, 1.9580, 2.2722, 2.1581, 1.9560, 1.1843], + device='cuda:1'), covar=tensor([0.5508, 0.4883, 0.1888, 0.3778, 0.2451, 0.3141, 0.1841, 0.5010], + device='cuda:1'), in_proj_covar=tensor([0.0938, 0.0970, 0.0795, 0.0936, 0.0987, 0.0882, 0.0741, 0.0820], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 01:08:04,192 INFO [train.py:935] (1/4) Epoch 20, validation: loss=0.175, simple_loss=0.275, pruned_loss=0.03755, over 944034.00 frames. +2023-02-07 01:08:04,193 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 01:08:09,555 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.504e+02 2.869e+02 3.482e+02 8.370e+02, threshold=5.739e+02, percent-clipped=5.0 +2023-02-07 01:08:15,887 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159593.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 01:08:38,920 INFO [train.py:901] (1/4) Epoch 20, batch 6050, loss[loss=0.2258, simple_loss=0.3012, pruned_loss=0.07521, over 7176.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2904, pruned_loss=0.06353, over 1612203.01 frames. ], batch size: 71, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:08:45,949 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159636.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:08:55,617 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159649.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:08:57,636 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159652.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:09:14,770 INFO [train.py:901] (1/4) Epoch 20, batch 6100, loss[loss=0.2612, simple_loss=0.3354, pruned_loss=0.0935, over 8621.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2905, pruned_loss=0.06389, over 1613089.69 frames. ], batch size: 31, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:09:21,004 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.453e+02 2.842e+02 3.745e+02 1.322e+03, threshold=5.684e+02, percent-clipped=4.0 +2023-02-07 01:09:34,330 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9836, 1.6774, 2.0712, 1.8207, 1.9688, 2.0108, 1.8423, 0.7858], + device='cuda:1'), covar=tensor([0.5386, 0.4540, 0.1886, 0.3375, 0.2243, 0.3012, 0.1863, 0.4913], + device='cuda:1'), in_proj_covar=tensor([0.0937, 0.0971, 0.0795, 0.0937, 0.0987, 0.0883, 0.0741, 0.0820], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 01:09:39,605 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6879, 4.7022, 4.2368, 1.8993, 4.2246, 4.3566, 4.2819, 4.1312], + device='cuda:1'), covar=tensor([0.0744, 0.0542, 0.0960, 0.5124, 0.0844, 0.0909, 0.1131, 0.0676], + device='cuda:1'), in_proj_covar=tensor([0.0519, 0.0430, 0.0434, 0.0533, 0.0422, 0.0435, 0.0418, 0.0377], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:09:41,577 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 01:09:50,003 INFO [train.py:901] (1/4) Epoch 20, batch 6150, loss[loss=0.1819, simple_loss=0.2745, pruned_loss=0.04462, over 8457.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2895, pruned_loss=0.06334, over 1610606.51 frames. ], batch size: 27, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:09:55,568 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0678, 2.3253, 1.9503, 2.8740, 1.3480, 1.6858, 2.0663, 2.2140], + device='cuda:1'), covar=tensor([0.0708, 0.0751, 0.0870, 0.0399, 0.1095, 0.1230, 0.0851, 0.0779], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0196, 0.0245, 0.0213, 0.0204, 0.0246, 0.0248, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 01:10:18,350 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159767.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:10:24,867 INFO [train.py:901] (1/4) Epoch 20, batch 6200, loss[loss=0.1886, simple_loss=0.2757, pruned_loss=0.05076, over 7971.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2893, pruned_loss=0.06324, over 1613830.88 frames. ], batch size: 21, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:10:30,198 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.429e+02 3.094e+02 3.753e+02 7.329e+02, threshold=6.188e+02, percent-clipped=3.0 +2023-02-07 01:10:43,487 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159802.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:11:00,325 INFO [train.py:901] (1/4) Epoch 20, batch 6250, loss[loss=0.1981, simple_loss=0.2778, pruned_loss=0.05925, over 8243.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2891, pruned_loss=0.06292, over 1615349.49 frames. ], batch size: 22, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:11:01,214 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159827.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:11:15,440 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-02-07 01:11:32,491 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159873.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:11:34,366 INFO [train.py:901] (1/4) Epoch 20, batch 6300, loss[loss=0.2084, simple_loss=0.3035, pruned_loss=0.05662, over 8242.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2883, pruned_loss=0.06256, over 1613625.94 frames. ], batch size: 24, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:11:40,336 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.354e+02 2.951e+02 3.644e+02 9.166e+02, threshold=5.902e+02, percent-clipped=5.0 +2023-02-07 01:11:45,865 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159892.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:12:03,509 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159917.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:12:09,078 INFO [train.py:901] (1/4) Epoch 20, batch 6350, loss[loss=0.2523, simple_loss=0.3258, pruned_loss=0.08946, over 8521.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2893, pruned_loss=0.06318, over 1618846.28 frames. ], batch size: 28, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:12:10,588 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159928.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:12:43,479 INFO [train.py:901] (1/4) Epoch 20, batch 6400, loss[loss=0.1802, simple_loss=0.2605, pruned_loss=0.04991, over 7807.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2873, pruned_loss=0.06191, over 1617604.11 frames. ], batch size: 20, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:12:48,765 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 2.436e+02 2.995e+02 3.881e+02 8.346e+02, threshold=5.989e+02, percent-clipped=6.0 +2023-02-07 01:12:55,755 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159993.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:13:16,776 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160023.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:13:18,631 INFO [train.py:901] (1/4) Epoch 20, batch 6450, loss[loss=0.2438, simple_loss=0.3303, pruned_loss=0.07868, over 8482.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2879, pruned_loss=0.06267, over 1613672.34 frames. ], batch size: 29, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:13:34,493 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160048.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:13:53,995 INFO [train.py:901] (1/4) Epoch 20, batch 6500, loss[loss=0.2382, simple_loss=0.3031, pruned_loss=0.08667, over 7934.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2876, pruned_loss=0.06259, over 1613013.53 frames. ], batch size: 20, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:13:59,465 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.613e+02 3.061e+02 4.120e+02 1.100e+03, threshold=6.122e+02, percent-clipped=8.0 +2023-02-07 01:14:16,515 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160108.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:14:25,335 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6178, 1.6087, 2.0990, 1.4679, 1.2296, 2.0327, 0.3719, 1.2786], + device='cuda:1'), covar=tensor([0.1878, 0.1334, 0.0362, 0.1102, 0.2744, 0.0457, 0.2162, 0.1360], + device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0195, 0.0126, 0.0221, 0.0271, 0.0133, 0.0169, 0.0190], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 01:14:29,725 INFO [train.py:901] (1/4) Epoch 20, batch 6550, loss[loss=0.2748, simple_loss=0.3444, pruned_loss=0.1026, over 8511.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2885, pruned_loss=0.06328, over 1612276.78 frames. ], batch size: 28, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:14:44,611 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 +2023-02-07 01:14:53,089 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 01:15:05,565 INFO [train.py:901] (1/4) Epoch 20, batch 6600, loss[loss=0.2198, simple_loss=0.3059, pruned_loss=0.06688, over 8310.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2885, pruned_loss=0.06294, over 1610502.30 frames. ], batch size: 49, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:15:10,795 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.488e+02 3.067e+02 3.982e+02 8.719e+02, threshold=6.134e+02, percent-clipped=3.0 +2023-02-07 01:15:12,114 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 01:15:13,063 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6960, 2.0987, 3.4200, 1.4903, 2.5419, 2.2232, 1.8328, 2.5444], + device='cuda:1'), covar=tensor([0.1794, 0.2710, 0.0787, 0.4393, 0.1748, 0.2998, 0.2187, 0.2204], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0594, 0.0557, 0.0637, 0.0645, 0.0597, 0.0533, 0.0635], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:15:13,752 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8082, 2.3598, 4.3282, 1.5925, 3.0103, 2.4602, 1.9460, 3.0230], + device='cuda:1'), covar=tensor([0.1856, 0.2799, 0.0724, 0.4473, 0.1974, 0.3096, 0.2225, 0.2391], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0594, 0.0557, 0.0637, 0.0645, 0.0597, 0.0533, 0.0635], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:15:22,710 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6788, 1.6018, 3.2057, 1.4295, 2.1904, 3.5456, 3.6453, 2.8666], + device='cuda:1'), covar=tensor([0.1452, 0.1844, 0.0442, 0.2312, 0.1187, 0.0302, 0.0641, 0.0774], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0321, 0.0287, 0.0315, 0.0306, 0.0263, 0.0412, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 01:15:29,491 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6244, 1.3589, 1.6104, 1.2588, 0.8489, 1.3389, 1.4445, 1.4127], + device='cuda:1'), covar=tensor([0.0603, 0.1367, 0.1701, 0.1529, 0.0585, 0.1550, 0.0756, 0.0633], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0153, 0.0190, 0.0159, 0.0099, 0.0162, 0.0112, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 01:15:33,423 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=160217.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:15:39,320 INFO [train.py:901] (1/4) Epoch 20, batch 6650, loss[loss=0.2279, simple_loss=0.3015, pruned_loss=0.07713, over 8435.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2888, pruned_loss=0.06352, over 1608956.45 frames. ], batch size: 27, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:15:49,669 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8378, 1.3263, 4.0149, 1.4734, 3.5670, 3.3401, 3.6761, 3.5526], + device='cuda:1'), covar=tensor([0.0679, 0.4599, 0.0590, 0.4148, 0.1149, 0.0970, 0.0637, 0.0721], + device='cuda:1'), in_proj_covar=tensor([0.0613, 0.0633, 0.0684, 0.0619, 0.0696, 0.0597, 0.0600, 0.0670], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:16:12,500 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=160272.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:16:15,124 INFO [train.py:901] (1/4) Epoch 20, batch 6700, loss[loss=0.1864, simple_loss=0.2528, pruned_loss=0.06001, over 7294.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2888, pruned_loss=0.06341, over 1612226.72 frames. ], batch size: 16, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:16:20,502 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.299e+02 2.819e+02 3.357e+02 8.975e+02, threshold=5.638e+02, percent-clipped=4.0 +2023-02-07 01:16:48,910 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160325.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:16:49,345 INFO [train.py:901] (1/4) Epoch 20, batch 6750, loss[loss=0.2347, simple_loss=0.2977, pruned_loss=0.0858, over 7650.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2887, pruned_loss=0.06316, over 1613175.73 frames. ], batch size: 19, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:16:53,607 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160332.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:17:06,675 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1341, 1.4364, 1.6327, 1.2735, 1.0474, 1.3628, 1.8624, 1.5404], + device='cuda:1'), covar=tensor([0.0537, 0.1287, 0.1680, 0.1517, 0.0603, 0.1523, 0.0695, 0.0650], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0152, 0.0190, 0.0159, 0.0099, 0.0162, 0.0113, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 01:17:06,882 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-07 01:17:16,338 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160364.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:17:23,986 INFO [train.py:901] (1/4) Epoch 20, batch 6800, loss[loss=0.2186, simple_loss=0.3088, pruned_loss=0.0642, over 8466.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2885, pruned_loss=0.06282, over 1612702.51 frames. ], batch size: 29, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:17:28,102 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 01:17:29,320 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.714e+02 2.510e+02 3.096e+02 3.947e+02 9.727e+02, threshold=6.192e+02, percent-clipped=5.0 +2023-02-07 01:17:31,586 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160387.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:17:33,656 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160389.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:17:37,188 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5980, 1.8047, 1.8741, 1.4192, 1.9778, 1.4336, 0.5625, 1.8220], + device='cuda:1'), covar=tensor([0.0447, 0.0309, 0.0269, 0.0417, 0.0340, 0.0757, 0.0782, 0.0234], + device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0376, 0.0331, 0.0432, 0.0362, 0.0523, 0.0383, 0.0403], + device='cuda:1'), out_proj_covar=tensor([1.1873e-04, 9.8639e-05, 8.7278e-05, 1.1426e-04, 9.5605e-05, 1.4867e-04, + 1.0356e-04, 1.0729e-04], device='cuda:1') +2023-02-07 01:17:59,211 INFO [train.py:901] (1/4) Epoch 20, batch 6850, loss[loss=0.1952, simple_loss=0.2663, pruned_loss=0.06202, over 7562.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2879, pruned_loss=0.06276, over 1606644.19 frames. ], batch size: 18, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:18:19,446 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 01:18:34,197 INFO [train.py:901] (1/4) Epoch 20, batch 6900, loss[loss=0.1851, simple_loss=0.268, pruned_loss=0.05112, over 8297.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2881, pruned_loss=0.06255, over 1609739.45 frames. ], batch size: 23, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:18:39,573 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.333e+02 2.912e+02 3.495e+02 9.213e+02, threshold=5.824e+02, percent-clipped=3.0 +2023-02-07 01:18:54,841 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4075, 2.2136, 3.1518, 2.4445, 3.0371, 2.3737, 2.1153, 1.8186], + device='cuda:1'), covar=tensor([0.5404, 0.5061, 0.1964, 0.3893, 0.2510, 0.3124, 0.1999, 0.5682], + device='cuda:1'), in_proj_covar=tensor([0.0929, 0.0968, 0.0794, 0.0932, 0.0985, 0.0881, 0.0740, 0.0821], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 01:19:02,431 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.89 vs. limit=5.0 +2023-02-07 01:19:08,574 INFO [train.py:901] (1/4) Epoch 20, batch 6950, loss[loss=0.2473, simple_loss=0.3213, pruned_loss=0.08669, over 8315.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2882, pruned_loss=0.06258, over 1612169.45 frames. ], batch size: 25, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:19:30,214 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 01:19:42,961 INFO [train.py:901] (1/4) Epoch 20, batch 7000, loss[loss=0.1934, simple_loss=0.2884, pruned_loss=0.04915, over 8444.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2877, pruned_loss=0.06209, over 1612933.37 frames. ], batch size: 29, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:19:48,348 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.497e+02 2.987e+02 3.377e+02 5.985e+02, threshold=5.974e+02, percent-clipped=1.0 +2023-02-07 01:19:49,093 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7249, 4.7261, 4.2225, 2.0765, 4.1761, 4.3761, 4.1925, 4.2385], + device='cuda:1'), covar=tensor([0.0708, 0.0505, 0.1022, 0.4404, 0.0876, 0.0914, 0.1319, 0.0651], + device='cuda:1'), in_proj_covar=tensor([0.0522, 0.0434, 0.0437, 0.0538, 0.0424, 0.0440, 0.0423, 0.0383], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:19:52,051 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6405, 1.6440, 2.1026, 1.4742, 1.2986, 2.0875, 0.3762, 1.3899], + device='cuda:1'), covar=tensor([0.1636, 0.1274, 0.0458, 0.1192, 0.2546, 0.0414, 0.2067, 0.1236], + device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0193, 0.0126, 0.0222, 0.0270, 0.0133, 0.0169, 0.0189], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 01:19:52,064 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160588.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:20:09,541 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160613.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:20:18,199 INFO [train.py:901] (1/4) Epoch 20, batch 7050, loss[loss=0.1993, simple_loss=0.2811, pruned_loss=0.05874, over 7820.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2888, pruned_loss=0.06275, over 1611475.47 frames. ], batch size: 20, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:20:30,580 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:20:42,512 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7724, 2.3782, 4.3599, 1.6153, 3.0422, 2.4281, 1.8657, 3.1339], + device='cuda:1'), covar=tensor([0.1779, 0.2412, 0.0722, 0.4146, 0.1765, 0.2891, 0.2117, 0.2087], + device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0585, 0.0548, 0.0627, 0.0636, 0.0587, 0.0523, 0.0623], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:20:48,791 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160668.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:20:49,342 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=160669.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:20:53,993 INFO [train.py:901] (1/4) Epoch 20, batch 7100, loss[loss=0.2024, simple_loss=0.2778, pruned_loss=0.06344, over 8499.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2884, pruned_loss=0.06284, over 1609278.69 frames. ], batch size: 26, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:20:59,624 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.520e+02 2.814e+02 3.523e+02 7.232e+02, threshold=5.628e+02, percent-clipped=2.0 +2023-02-07 01:21:00,543 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4511, 2.8075, 2.3171, 3.8427, 1.7301, 2.1624, 2.3391, 2.8618], + device='cuda:1'), covar=tensor([0.0693, 0.0870, 0.0870, 0.0294, 0.1156, 0.1166, 0.1025, 0.0820], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0198, 0.0247, 0.0214, 0.0206, 0.0248, 0.0252, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 01:21:29,424 INFO [train.py:901] (1/4) Epoch 20, batch 7150, loss[loss=0.1885, simple_loss=0.2698, pruned_loss=0.05358, over 7821.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2894, pruned_loss=0.06333, over 1608834.48 frames. ], batch size: 20, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:21:36,643 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.9112, 3.8550, 3.5190, 1.9815, 3.4632, 3.5284, 3.3619, 3.4026], + device='cuda:1'), covar=tensor([0.0827, 0.0601, 0.1131, 0.3994, 0.0933, 0.0856, 0.1477, 0.0747], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0432, 0.0436, 0.0536, 0.0424, 0.0440, 0.0424, 0.0381], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:21:54,043 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 01:22:04,606 INFO [train.py:901] (1/4) Epoch 20, batch 7200, loss[loss=0.2559, simple_loss=0.3193, pruned_loss=0.09622, over 6933.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.29, pruned_loss=0.06336, over 1612964.91 frames. ], batch size: 71, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:22:09,775 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.432e+02 3.066e+02 3.972e+02 8.502e+02, threshold=6.132e+02, percent-clipped=3.0 +2023-02-07 01:22:09,970 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160784.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 01:22:26,159 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 01:22:39,218 INFO [train.py:901] (1/4) Epoch 20, batch 7250, loss[loss=0.1921, simple_loss=0.2649, pruned_loss=0.05963, over 7665.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2899, pruned_loss=0.06368, over 1610461.34 frames. ], batch size: 19, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:23:13,942 INFO [train.py:901] (1/4) Epoch 20, batch 7300, loss[loss=0.1939, simple_loss=0.2789, pruned_loss=0.0544, over 7816.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2896, pruned_loss=0.06304, over 1612882.69 frames. ], batch size: 20, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:23:14,055 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3970, 4.3968, 3.9466, 2.0179, 3.9013, 4.0243, 3.9752, 3.8163], + device='cuda:1'), covar=tensor([0.0711, 0.0513, 0.1067, 0.4523, 0.0838, 0.0920, 0.1178, 0.0780], + device='cuda:1'), in_proj_covar=tensor([0.0519, 0.0431, 0.0434, 0.0536, 0.0426, 0.0439, 0.0421, 0.0379], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:23:19,318 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.519e+02 2.885e+02 3.982e+02 8.183e+02, threshold=5.771e+02, percent-clipped=5.0 +2023-02-07 01:23:44,252 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160919.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:23:48,786 INFO [train.py:901] (1/4) Epoch 20, batch 7350, loss[loss=0.2011, simple_loss=0.283, pruned_loss=0.05954, over 8345.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2889, pruned_loss=0.06274, over 1611685.12 frames. ], batch size: 24, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:24:07,563 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0879, 2.2081, 1.8321, 2.7579, 1.2403, 1.5587, 1.9021, 2.1783], + device='cuda:1'), covar=tensor([0.0639, 0.0712, 0.0905, 0.0370, 0.1170, 0.1336, 0.0883, 0.0684], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0198, 0.0246, 0.0213, 0.0206, 0.0247, 0.0251, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 01:24:16,156 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 01:24:24,317 INFO [train.py:901] (1/4) Epoch 20, batch 7400, loss[loss=0.1885, simple_loss=0.2815, pruned_loss=0.04773, over 8075.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2883, pruned_loss=0.06232, over 1615182.50 frames. ], batch size: 21, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:24:29,923 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160983.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:24:30,420 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.344e+02 3.002e+02 3.673e+02 6.079e+02, threshold=6.004e+02, percent-clipped=1.0 +2023-02-07 01:24:31,262 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7842, 1.4111, 3.1222, 1.3108, 2.2813, 3.3409, 3.4816, 2.7971], + device='cuda:1'), covar=tensor([0.1274, 0.1882, 0.0375, 0.2298, 0.0916, 0.0273, 0.0642, 0.0622], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0321, 0.0285, 0.0315, 0.0305, 0.0262, 0.0412, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 01:24:37,302 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 01:24:59,972 INFO [train.py:901] (1/4) Epoch 20, batch 7450, loss[loss=0.1824, simple_loss=0.2686, pruned_loss=0.0481, over 7930.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2886, pruned_loss=0.0624, over 1616522.21 frames. ], batch size: 20, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:25:10,069 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161040.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 01:25:16,121 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 01:25:28,631 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161065.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:25:35,932 INFO [train.py:901] (1/4) Epoch 20, batch 7500, loss[loss=0.1797, simple_loss=0.258, pruned_loss=0.05074, over 8253.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2885, pruned_loss=0.06276, over 1616291.83 frames. ], batch size: 22, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:25:41,427 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.441e+02 3.010e+02 3.756e+02 8.900e+02, threshold=6.020e+02, percent-clipped=5.0 +2023-02-07 01:25:56,410 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7495, 1.9671, 2.0457, 1.3868, 2.2003, 1.6378, 0.6352, 1.8879], + device='cuda:1'), covar=tensor([0.0567, 0.0334, 0.0279, 0.0539, 0.0381, 0.0846, 0.0893, 0.0304], + device='cuda:1'), in_proj_covar=tensor([0.0451, 0.0385, 0.0339, 0.0440, 0.0370, 0.0534, 0.0391, 0.0413], + device='cuda:1'), out_proj_covar=tensor([1.2137e-04, 1.0108e-04, 8.9232e-05, 1.1615e-04, 9.7755e-05, 1.5176e-04, + 1.0577e-04, 1.0976e-04], device='cuda:1') +2023-02-07 01:26:11,154 INFO [train.py:901] (1/4) Epoch 20, batch 7550, loss[loss=0.2073, simple_loss=0.2897, pruned_loss=0.06241, over 7977.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2893, pruned_loss=0.06341, over 1615033.51 frames. ], batch size: 21, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:26:32,389 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7714, 1.7436, 2.3980, 1.6107, 1.4213, 2.4572, 0.4663, 1.4671], + device='cuda:1'), covar=tensor([0.1827, 0.1233, 0.0376, 0.1326, 0.2655, 0.0413, 0.2301, 0.1352], + device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0194, 0.0125, 0.0220, 0.0269, 0.0134, 0.0168, 0.0188], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 01:26:38,652 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.61 vs. limit=5.0 +2023-02-07 01:26:40,487 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5664, 1.8237, 1.8682, 1.2280, 1.9856, 1.4964, 0.4245, 1.7330], + device='cuda:1'), covar=tensor([0.0462, 0.0312, 0.0270, 0.0482, 0.0361, 0.0765, 0.0791, 0.0245], + device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0384, 0.0338, 0.0439, 0.0368, 0.0533, 0.0390, 0.0412], + device='cuda:1'), out_proj_covar=tensor([1.2109e-04, 1.0080e-04, 8.9035e-05, 1.1583e-04, 9.7149e-05, 1.5137e-04, + 1.0544e-04, 1.0964e-04], device='cuda:1') +2023-02-07 01:26:46,340 INFO [train.py:901] (1/4) Epoch 20, batch 7600, loss[loss=0.1997, simple_loss=0.2872, pruned_loss=0.05611, over 8332.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2905, pruned_loss=0.0638, over 1618708.02 frames. ], batch size: 25, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:26:51,651 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.460e+02 3.037e+02 4.113e+02 9.859e+02, threshold=6.074e+02, percent-clipped=9.0 +2023-02-07 01:26:56,504 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 01:27:20,297 INFO [train.py:901] (1/4) Epoch 20, batch 7650, loss[loss=0.1828, simple_loss=0.2647, pruned_loss=0.05043, over 7789.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2902, pruned_loss=0.06387, over 1616837.48 frames. ], batch size: 19, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:27:25,750 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161234.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:27:36,297 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161249.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:27:45,610 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161263.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:27:54,281 INFO [train.py:901] (1/4) Epoch 20, batch 7700, loss[loss=0.1743, simple_loss=0.2619, pruned_loss=0.04339, over 7689.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2888, pruned_loss=0.06305, over 1609729.20 frames. ], batch size: 18, lr: 3.75e-03, grad_scale: 16.0 +2023-02-07 01:27:59,457 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.411e+02 2.987e+02 3.572e+02 6.786e+02, threshold=5.975e+02, percent-clipped=3.0 +2023-02-07 01:28:25,738 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 01:28:29,788 INFO [train.py:901] (1/4) Epoch 20, batch 7750, loss[loss=0.1918, simple_loss=0.2802, pruned_loss=0.05177, over 8522.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2889, pruned_loss=0.06352, over 1607746.72 frames. ], batch size: 28, lr: 3.75e-03, grad_scale: 16.0 +2023-02-07 01:28:30,573 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161327.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:29:01,925 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5804, 4.6555, 4.1342, 2.0979, 4.1020, 4.1519, 4.1062, 3.9446], + device='cuda:1'), covar=tensor([0.0681, 0.0475, 0.1111, 0.4193, 0.0835, 0.0777, 0.1250, 0.0639], + device='cuda:1'), in_proj_covar=tensor([0.0510, 0.0425, 0.0428, 0.0526, 0.0419, 0.0429, 0.0413, 0.0375], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:29:05,270 INFO [train.py:901] (1/4) Epoch 20, batch 7800, loss[loss=0.1778, simple_loss=0.256, pruned_loss=0.04986, over 7213.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2884, pruned_loss=0.06287, over 1609897.37 frames. ], batch size: 16, lr: 3.75e-03, grad_scale: 16.0 +2023-02-07 01:29:06,858 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161378.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:29:10,615 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.672e+02 2.429e+02 2.909e+02 3.732e+02 6.331e+02, threshold=5.818e+02, percent-clipped=2.0 +2023-02-07 01:29:34,441 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161419.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:29:39,043 INFO [train.py:901] (1/4) Epoch 20, batch 7850, loss[loss=0.1845, simple_loss=0.2631, pruned_loss=0.05292, over 7932.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.289, pruned_loss=0.06326, over 1609049.93 frames. ], batch size: 20, lr: 3.74e-03, grad_scale: 16.0 +2023-02-07 01:29:49,655 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161442.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:29:57,456 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1328, 1.8162, 2.0249, 1.8580, 0.9214, 1.8221, 2.3480, 2.2843], + device='cuda:1'), covar=tensor([0.0407, 0.1205, 0.1550, 0.1336, 0.0582, 0.1402, 0.0568, 0.0577], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0158, 0.0099, 0.0161, 0.0112, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 01:30:12,445 INFO [train.py:901] (1/4) Epoch 20, batch 7900, loss[loss=0.182, simple_loss=0.262, pruned_loss=0.05104, over 7221.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2898, pruned_loss=0.06364, over 1602986.56 frames. ], batch size: 16, lr: 3.74e-03, grad_scale: 8.0 +2023-02-07 01:30:13,738 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161478.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:30:18,874 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 2.352e+02 2.923e+02 4.060e+02 8.940e+02, threshold=5.846e+02, percent-clipped=3.0 +2023-02-07 01:30:45,503 INFO [train.py:901] (1/4) Epoch 20, batch 7950, loss[loss=0.2065, simple_loss=0.2976, pruned_loss=0.05767, over 8315.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2911, pruned_loss=0.06456, over 1599689.59 frames. ], batch size: 25, lr: 3.74e-03, grad_scale: 8.0 +2023-02-07 01:30:50,382 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-07 01:30:51,104 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-07 01:31:18,038 INFO [train.py:901] (1/4) Epoch 20, batch 8000, loss[loss=0.1898, simple_loss=0.2664, pruned_loss=0.05664, over 7411.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2905, pruned_loss=0.06437, over 1599207.11 frames. ], batch size: 17, lr: 3.74e-03, grad_scale: 8.0 +2023-02-07 01:31:19,441 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161578.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:31:23,850 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.449e+02 3.108e+02 3.740e+02 8.675e+02, threshold=6.215e+02, percent-clipped=6.0 +2023-02-07 01:31:24,608 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0013, 1.3943, 3.4231, 1.5444, 2.3082, 3.8087, 3.9369, 3.1811], + device='cuda:1'), covar=tensor([0.1147, 0.1871, 0.0334, 0.2114, 0.1079, 0.0223, 0.0483, 0.0562], + device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0320, 0.0285, 0.0314, 0.0305, 0.0260, 0.0410, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 01:31:29,407 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161593.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:31:44,061 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8036, 2.7712, 2.0004, 2.4949, 2.4120, 1.6944, 2.3722, 2.5161], + device='cuda:1'), covar=tensor([0.1255, 0.0346, 0.1080, 0.0520, 0.0582, 0.1413, 0.0793, 0.0748], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0234, 0.0331, 0.0303, 0.0298, 0.0332, 0.0341, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 01:31:51,317 INFO [train.py:901] (1/4) Epoch 20, batch 8050, loss[loss=0.2854, simple_loss=0.3419, pruned_loss=0.1144, over 7208.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2887, pruned_loss=0.06387, over 1590174.81 frames. ], batch size: 72, lr: 3.74e-03, grad_scale: 8.0 +2023-02-07 01:31:52,873 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3738, 1.7739, 1.9312, 1.7554, 1.2565, 1.7834, 2.1964, 1.9449], + device='cuda:1'), covar=tensor([0.0619, 0.1424, 0.1914, 0.1573, 0.0777, 0.1650, 0.0762, 0.0643], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0158, 0.0099, 0.0162, 0.0112, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 01:31:57,072 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161634.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:32:11,404 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0523, 1.2686, 1.6669, 1.0051, 1.1753, 1.2353, 1.1299, 1.2039], + device='cuda:1'), covar=tensor([0.1400, 0.1797, 0.0679, 0.3128, 0.1448, 0.2415, 0.1632, 0.1897], + device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0598, 0.0554, 0.0641, 0.0647, 0.0597, 0.0531, 0.0636], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:32:12,862 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-07 01:32:24,914 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 01:32:32,185 INFO [train.py:901] (1/4) Epoch 21, batch 0, loss[loss=0.1866, simple_loss=0.269, pruned_loss=0.05215, over 7918.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.269, pruned_loss=0.05215, over 7918.00 frames. ], batch size: 20, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:32:32,186 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 01:32:44,214 INFO [train.py:935] (1/4) Epoch 21, validation: loss=0.1763, simple_loss=0.2762, pruned_loss=0.03818, over 944034.00 frames. +2023-02-07 01:32:44,214 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 01:32:44,412 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161659.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:32:59,352 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 01:33:02,217 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.415e+02 2.918e+02 3.924e+02 7.413e+02, threshold=5.835e+02, percent-clipped=4.0 +2023-02-07 01:33:07,958 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161693.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:33:11,595 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161698.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:33:18,547 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6619, 1.3796, 4.8766, 1.8036, 4.2204, 4.0884, 4.4745, 4.3148], + device='cuda:1'), covar=tensor([0.0629, 0.4725, 0.0450, 0.3924, 0.1152, 0.0978, 0.0556, 0.0615], + device='cuda:1'), in_proj_covar=tensor([0.0610, 0.0623, 0.0673, 0.0608, 0.0690, 0.0590, 0.0591, 0.0656], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:33:18,592 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161708.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:33:19,110 INFO [train.py:901] (1/4) Epoch 21, batch 50, loss[loss=0.24, simple_loss=0.3111, pruned_loss=0.08447, over 8107.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2906, pruned_loss=0.06324, over 368507.23 frames. ], batch size: 23, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:33:29,241 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:33:32,472 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 01:33:56,008 INFO [train.py:901] (1/4) Epoch 21, batch 100, loss[loss=0.2066, simple_loss=0.2941, pruned_loss=0.05951, over 8507.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2882, pruned_loss=0.06145, over 644126.92 frames. ], batch size: 26, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:33:57,263 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 01:33:58,657 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161763.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:34:14,151 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.511e+02 2.964e+02 4.065e+02 7.207e+02, threshold=5.927e+02, percent-clipped=4.0 +2023-02-07 01:34:18,656 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 01:34:30,769 INFO [train.py:901] (1/4) Epoch 21, batch 150, loss[loss=0.2043, simple_loss=0.2812, pruned_loss=0.06369, over 7416.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2878, pruned_loss=0.06161, over 861123.39 frames. ], batch size: 17, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:34:33,278 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 01:34:39,728 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161822.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:34:47,237 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161833.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:35:06,328 INFO [train.py:901] (1/4) Epoch 21, batch 200, loss[loss=0.1901, simple_loss=0.2743, pruned_loss=0.05295, over 8478.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2903, pruned_loss=0.06342, over 1024129.77 frames. ], batch size: 48, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:35:19,123 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161878.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:35:23,721 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.494e+02 2.791e+02 3.613e+02 7.338e+02, threshold=5.582e+02, percent-clipped=1.0 +2023-02-07 01:35:41,064 INFO [train.py:901] (1/4) Epoch 21, batch 250, loss[loss=0.181, simple_loss=0.2612, pruned_loss=0.0504, over 7806.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2905, pruned_loss=0.06396, over 1154648.11 frames. ], batch size: 20, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:35:47,956 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 01:35:57,090 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 01:36:00,653 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161937.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:36:08,767 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161949.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:36:15,274 INFO [train.py:901] (1/4) Epoch 21, batch 300, loss[loss=0.2169, simple_loss=0.2796, pruned_loss=0.07713, over 7439.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2894, pruned_loss=0.06371, over 1253212.27 frames. ], batch size: 17, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:36:19,038 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161964.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:36:20,643 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 01:36:26,608 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161974.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:36:33,755 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.407e+02 2.839e+02 3.558e+02 8.067e+02, threshold=5.678e+02, percent-clipped=5.0 +2023-02-07 01:36:36,650 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161989.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:36:51,865 INFO [train.py:901] (1/4) Epoch 21, batch 350, loss[loss=0.2075, simple_loss=0.2783, pruned_loss=0.06832, over 7811.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2892, pruned_loss=0.06313, over 1334308.22 frames. ], batch size: 20, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:36:57,888 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-07 01:37:25,811 INFO [train.py:901] (1/4) Epoch 21, batch 400, loss[loss=0.2291, simple_loss=0.3111, pruned_loss=0.07357, over 8499.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2902, pruned_loss=0.06371, over 1396037.14 frames. ], batch size: 26, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:37:44,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.323e+02 2.796e+02 3.394e+02 5.024e+02, threshold=5.592e+02, percent-clipped=0.0 +2023-02-07 01:37:52,936 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162095.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:38:02,164 INFO [train.py:901] (1/4) Epoch 21, batch 450, loss[loss=0.2368, simple_loss=0.3254, pruned_loss=0.07414, over 8240.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.29, pruned_loss=0.06374, over 1445798.62 frames. ], batch size: 24, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:38:15,018 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-07 01:38:20,151 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162134.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:38:37,311 INFO [train.py:901] (1/4) Epoch 21, batch 500, loss[loss=0.2072, simple_loss=0.278, pruned_loss=0.06819, over 8052.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2892, pruned_loss=0.06297, over 1485783.54 frames. ], batch size: 20, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:38:37,535 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162159.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:38:50,066 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=162177.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:38:55,585 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.501e+02 2.975e+02 3.750e+02 9.376e+02, threshold=5.950e+02, percent-clipped=8.0 +2023-02-07 01:39:01,450 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162193.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:39:13,738 INFO [train.py:901] (1/4) Epoch 21, batch 550, loss[loss=0.2158, simple_loss=0.2923, pruned_loss=0.06959, over 7712.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2895, pruned_loss=0.06313, over 1514896.82 frames. ], batch size: 18, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:39:20,129 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162218.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:39:36,008 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8063, 3.7677, 3.4095, 1.6954, 3.3030, 3.4250, 3.4011, 3.2552], + device='cuda:1'), covar=tensor([0.0967, 0.0703, 0.1205, 0.5260, 0.1068, 0.1203, 0.1325, 0.1037], + device='cuda:1'), in_proj_covar=tensor([0.0515, 0.0432, 0.0431, 0.0532, 0.0421, 0.0435, 0.0414, 0.0377], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:39:42,289 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162249.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:39:48,779 INFO [train.py:901] (1/4) Epoch 21, batch 600, loss[loss=0.1859, simple_loss=0.2645, pruned_loss=0.05367, over 7435.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2904, pruned_loss=0.06383, over 1534661.02 frames. ], batch size: 17, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:39:49,641 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7476, 2.4449, 4.9218, 2.9046, 4.5486, 4.2957, 4.6286, 4.4998], + device='cuda:1'), covar=tensor([0.0552, 0.3344, 0.0465, 0.2846, 0.0748, 0.0793, 0.0469, 0.0499], + device='cuda:1'), in_proj_covar=tensor([0.0618, 0.0635, 0.0684, 0.0616, 0.0700, 0.0601, 0.0600, 0.0667], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:40:02,415 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 01:40:06,579 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.365e+02 2.932e+02 3.412e+02 7.385e+02, threshold=5.863e+02, percent-clipped=2.0 +2023-02-07 01:40:08,822 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5485, 1.8054, 1.8427, 1.2837, 1.9489, 1.4426, 0.5354, 1.6976], + device='cuda:1'), covar=tensor([0.0532, 0.0311, 0.0271, 0.0493, 0.0323, 0.0723, 0.0816, 0.0271], + device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0383, 0.0335, 0.0437, 0.0367, 0.0530, 0.0387, 0.0411], + device='cuda:1'), out_proj_covar=tensor([1.2021e-04, 1.0044e-04, 8.8259e-05, 1.1543e-04, 9.6905e-05, 1.5057e-04, + 1.0474e-04, 1.0932e-04], device='cuda:1') +2023-02-07 01:40:11,428 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162292.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:40:22,965 INFO [train.py:901] (1/4) Epoch 21, batch 650, loss[loss=0.2288, simple_loss=0.311, pruned_loss=0.07332, over 8500.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2912, pruned_loss=0.0642, over 1552840.57 frames. ], batch size: 26, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:40:46,286 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4592, 2.3873, 1.7920, 2.1024, 2.0166, 1.5235, 1.9463, 1.8863], + device='cuda:1'), covar=tensor([0.1476, 0.0440, 0.1219, 0.0581, 0.0730, 0.1515, 0.0910, 0.0975], + device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0236, 0.0334, 0.0307, 0.0300, 0.0334, 0.0345, 0.0319], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 01:40:59,244 INFO [train.py:901] (1/4) Epoch 21, batch 700, loss[loss=0.1878, simple_loss=0.2724, pruned_loss=0.05163, over 7974.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2909, pruned_loss=0.06392, over 1564656.87 frames. ], batch size: 21, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:41:17,770 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 2.443e+02 3.111e+02 4.032e+02 8.821e+02, threshold=6.222e+02, percent-clipped=5.0 +2023-02-07 01:41:34,567 INFO [train.py:901] (1/4) Epoch 21, batch 750, loss[loss=0.2023, simple_loss=0.2879, pruned_loss=0.05831, over 8342.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2914, pruned_loss=0.06404, over 1576574.40 frames. ], batch size: 49, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:41:40,440 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 01:41:45,483 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 01:41:54,317 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 01:41:56,360 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=162439.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:42:01,037 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-07 01:42:11,066 INFO [train.py:901] (1/4) Epoch 21, batch 800, loss[loss=0.1836, simple_loss=0.2544, pruned_loss=0.05641, over 7420.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2901, pruned_loss=0.06374, over 1584029.10 frames. ], batch size: 17, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:42:29,940 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.455e+02 2.861e+02 3.570e+02 7.084e+02, threshold=5.721e+02, percent-clipped=3.0 +2023-02-07 01:42:47,171 INFO [train.py:901] (1/4) Epoch 21, batch 850, loss[loss=0.1927, simple_loss=0.279, pruned_loss=0.05323, over 8621.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2901, pruned_loss=0.06406, over 1586799.06 frames. ], batch size: 39, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:43:01,479 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162529.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:43:07,194 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5097, 2.4757, 3.3299, 2.6153, 3.0505, 2.6017, 2.3284, 1.8306], + device='cuda:1'), covar=tensor([0.5126, 0.4851, 0.1714, 0.3380, 0.2484, 0.2662, 0.1721, 0.5244], + device='cuda:1'), in_proj_covar=tensor([0.0936, 0.0967, 0.0791, 0.0929, 0.0983, 0.0878, 0.0738, 0.0815], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 01:43:16,271 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162548.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:43:21,105 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162554.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:43:24,457 INFO [train.py:901] (1/4) Epoch 21, batch 900, loss[loss=0.2296, simple_loss=0.3167, pruned_loss=0.07121, over 8124.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2888, pruned_loss=0.06323, over 1589734.98 frames. ], batch size: 22, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:43:34,386 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162573.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:43:42,642 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.319e+02 2.838e+02 3.637e+02 1.203e+03, threshold=5.677e+02, percent-clipped=5.0 +2023-02-07 01:43:49,232 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=162593.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:44:00,582 INFO [train.py:901] (1/4) Epoch 21, batch 950, loss[loss=0.2108, simple_loss=0.3061, pruned_loss=0.0577, over 8493.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2891, pruned_loss=0.06328, over 1597160.14 frames. ], batch size: 26, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:44:07,071 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8173, 1.6248, 2.5172, 1.9346, 2.1835, 1.8154, 1.5946, 1.1127], + device='cuda:1'), covar=tensor([0.6897, 0.5783, 0.1702, 0.3668, 0.2719, 0.4084, 0.2984, 0.4738], + device='cuda:1'), in_proj_covar=tensor([0.0937, 0.0969, 0.0791, 0.0930, 0.0983, 0.0878, 0.0739, 0.0817], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 01:44:14,244 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 01:44:18,634 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2966, 2.0639, 2.8074, 2.2732, 2.6712, 2.3147, 2.1124, 1.4714], + device='cuda:1'), covar=tensor([0.5165, 0.4686, 0.1845, 0.3608, 0.2430, 0.2876, 0.1820, 0.5161], + device='cuda:1'), in_proj_covar=tensor([0.0939, 0.0970, 0.0793, 0.0931, 0.0985, 0.0879, 0.0740, 0.0818], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 01:44:35,852 INFO [train.py:901] (1/4) Epoch 21, batch 1000, loss[loss=0.2407, simple_loss=0.3201, pruned_loss=0.08064, over 8480.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2899, pruned_loss=0.06385, over 1597597.76 frames. ], batch size: 25, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:44:48,965 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 01:44:55,195 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.438e+02 2.954e+02 4.014e+02 9.557e+02, threshold=5.908e+02, percent-clipped=3.0 +2023-02-07 01:45:01,392 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 01:45:11,699 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162708.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:45:12,214 INFO [train.py:901] (1/4) Epoch 21, batch 1050, loss[loss=0.2485, simple_loss=0.3198, pruned_loss=0.08858, over 8467.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2916, pruned_loss=0.06444, over 1608348.86 frames. ], batch size: 27, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:45:46,485 INFO [train.py:901] (1/4) Epoch 21, batch 1100, loss[loss=0.2058, simple_loss=0.2835, pruned_loss=0.06408, over 7930.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2922, pruned_loss=0.065, over 1614028.50 frames. ], batch size: 20, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:46:06,010 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.501e+02 3.059e+02 3.494e+02 1.150e+03, threshold=6.119e+02, percent-clipped=4.0 +2023-02-07 01:46:14,532 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 01:46:23,780 INFO [train.py:901] (1/4) Epoch 21, batch 1150, loss[loss=0.1926, simple_loss=0.2628, pruned_loss=0.0612, over 7697.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2915, pruned_loss=0.06401, over 1617108.21 frames. ], batch size: 18, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:46:24,670 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162810.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:46:29,071 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 01:46:29,779 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 01:46:43,000 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162835.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:46:50,079 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162845.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:46:59,714 INFO [train.py:901] (1/4) Epoch 21, batch 1200, loss[loss=0.2167, simple_loss=0.3025, pruned_loss=0.06545, over 8334.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2912, pruned_loss=0.0638, over 1615941.34 frames. ], batch size: 25, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:47:09,544 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=162873.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:47:17,453 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.784e+02 2.368e+02 3.051e+02 3.779e+02 6.869e+02, threshold=6.103e+02, percent-clipped=3.0 +2023-02-07 01:47:36,400 INFO [train.py:901] (1/4) Epoch 21, batch 1250, loss[loss=0.2378, simple_loss=0.3176, pruned_loss=0.079, over 8607.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2909, pruned_loss=0.06365, over 1614703.83 frames. ], batch size: 39, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:48:11,274 INFO [train.py:901] (1/4) Epoch 21, batch 1300, loss[loss=0.1552, simple_loss=0.2284, pruned_loss=0.04105, over 7722.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2895, pruned_loss=0.06265, over 1614834.62 frames. ], batch size: 18, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:48:14,772 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162964.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:48:27,753 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-07 01:48:28,485 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.260e+02 2.727e+02 3.317e+02 5.773e+02, threshold=5.453e+02, percent-clipped=0.0 +2023-02-07 01:48:29,597 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.85 vs. limit=5.0 +2023-02-07 01:48:30,726 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162988.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:48:31,444 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162989.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:48:44,797 INFO [train.py:901] (1/4) Epoch 21, batch 1350, loss[loss=0.198, simple_loss=0.2877, pruned_loss=0.05413, over 8557.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.289, pruned_loss=0.06244, over 1612079.22 frames. ], batch size: 31, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:48:45,178 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-07 01:49:09,955 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-07 01:49:21,723 INFO [train.py:901] (1/4) Epoch 21, batch 1400, loss[loss=0.1791, simple_loss=0.2548, pruned_loss=0.05165, over 7514.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2879, pruned_loss=0.06238, over 1611431.78 frames. ], batch size: 18, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:49:34,612 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 01:49:38,380 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.26 vs. limit=5.0 +2023-02-07 01:49:39,404 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.469e+02 3.010e+02 4.050e+02 1.060e+03, threshold=6.020e+02, percent-clipped=5.0 +2023-02-07 01:49:46,325 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 01:49:55,428 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163108.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:49:55,918 INFO [train.py:901] (1/4) Epoch 21, batch 1450, loss[loss=0.2259, simple_loss=0.3028, pruned_loss=0.07454, over 8468.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2891, pruned_loss=0.06287, over 1613903.20 frames. ], batch size: 25, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:50:31,010 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6535, 1.6616, 2.3340, 1.5078, 1.2092, 2.3184, 0.4340, 1.3796], + device='cuda:1'), covar=tensor([0.1836, 0.1368, 0.0304, 0.1337, 0.3010, 0.0390, 0.2382, 0.1489], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0192, 0.0125, 0.0219, 0.0269, 0.0132, 0.0167, 0.0188], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 01:50:32,141 INFO [train.py:901] (1/4) Epoch 21, batch 1500, loss[loss=0.2959, simple_loss=0.3482, pruned_loss=0.1218, over 7053.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2905, pruned_loss=0.06334, over 1616303.25 frames. ], batch size: 71, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:50:35,022 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1766, 2.5457, 3.0427, 1.6958, 3.3030, 1.8323, 1.5441, 2.0140], + device='cuda:1'), covar=tensor([0.0858, 0.0468, 0.0269, 0.0814, 0.0406, 0.0978, 0.1047, 0.0604], + device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0389, 0.0337, 0.0440, 0.0371, 0.0534, 0.0389, 0.0418], + device='cuda:1'), out_proj_covar=tensor([1.2093e-04, 1.0215e-04, 8.8828e-05, 1.1620e-04, 9.7726e-05, 1.5150e-04, + 1.0536e-04, 1.1126e-04], device='cuda:1') +2023-02-07 01:50:50,508 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.589e+02 2.250e+02 2.722e+02 3.392e+02 6.898e+02, threshold=5.444e+02, percent-clipped=4.0 +2023-02-07 01:50:53,289 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=163189.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:51:05,622 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9797, 2.0256, 1.7855, 2.6136, 1.2866, 1.6323, 1.9283, 2.0077], + device='cuda:1'), covar=tensor([0.0673, 0.0776, 0.0927, 0.0405, 0.1042, 0.1200, 0.0763, 0.0757], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0196, 0.0245, 0.0213, 0.0205, 0.0247, 0.0250, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 01:51:06,797 INFO [train.py:901] (1/4) Epoch 21, batch 1550, loss[loss=0.1762, simple_loss=0.2694, pruned_loss=0.04154, over 8108.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2892, pruned_loss=0.06298, over 1614994.12 frames. ], batch size: 23, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:51:31,325 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163244.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:51:42,279 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163258.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:51:42,818 INFO [train.py:901] (1/4) Epoch 21, batch 1600, loss[loss=0.1918, simple_loss=0.2718, pruned_loss=0.05584, over 8247.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2904, pruned_loss=0.06313, over 1619033.07 frames. ], batch size: 24, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:51:47,157 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5987, 1.4582, 1.6169, 1.3236, 0.9413, 1.4090, 1.4608, 1.2884], + device='cuda:1'), covar=tensor([0.0566, 0.1246, 0.1746, 0.1467, 0.0620, 0.1500, 0.0683, 0.0689], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0099, 0.0162, 0.0112, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 01:51:50,627 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163269.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:52:00,883 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.380e+02 3.009e+02 4.081e+02 9.131e+02, threshold=6.018e+02, percent-clipped=6.0 +2023-02-07 01:52:14,551 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163304.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:52:17,724 INFO [train.py:901] (1/4) Epoch 21, batch 1650, loss[loss=0.2236, simple_loss=0.3039, pruned_loss=0.07161, over 8436.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2903, pruned_loss=0.0634, over 1623061.84 frames. ], batch size: 39, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:52:51,352 INFO [train.py:901] (1/4) Epoch 21, batch 1700, loss[loss=0.2351, simple_loss=0.3043, pruned_loss=0.08297, over 6890.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2917, pruned_loss=0.06364, over 1626163.18 frames. ], batch size: 72, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:53:09,964 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.344e+02 2.897e+02 3.678e+02 1.033e+03, threshold=5.793e+02, percent-clipped=5.0 +2023-02-07 01:53:27,422 INFO [train.py:901] (1/4) Epoch 21, batch 1750, loss[loss=0.1966, simple_loss=0.2807, pruned_loss=0.05627, over 7923.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2909, pruned_loss=0.06326, over 1623351.12 frames. ], batch size: 20, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:53:51,719 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1269, 1.8760, 2.1428, 1.8199, 1.3427, 1.7435, 2.3606, 2.3052], + device='cuda:1'), covar=tensor([0.0375, 0.1113, 0.1486, 0.1305, 0.0539, 0.1362, 0.0572, 0.0551], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0188, 0.0158, 0.0098, 0.0161, 0.0112, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 01:53:56,338 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=163452.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:54:01,071 INFO [train.py:901] (1/4) Epoch 21, batch 1800, loss[loss=0.1993, simple_loss=0.287, pruned_loss=0.05585, over 8461.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2901, pruned_loss=0.06292, over 1618309.03 frames. ], batch size: 25, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:54:18,725 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.661e+02 3.025e+02 4.067e+02 7.408e+02, threshold=6.049e+02, percent-clipped=6.0 +2023-02-07 01:54:36,582 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 +2023-02-07 01:54:37,335 INFO [train.py:901] (1/4) Epoch 21, batch 1850, loss[loss=0.2266, simple_loss=0.3023, pruned_loss=0.07551, over 8433.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2899, pruned_loss=0.0632, over 1614784.86 frames. ], batch size: 27, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:55:11,686 INFO [train.py:901] (1/4) Epoch 21, batch 1900, loss[loss=0.2308, simple_loss=0.3167, pruned_loss=0.07246, over 8323.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.29, pruned_loss=0.06341, over 1616512.88 frames. ], batch size: 26, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:55:12,592 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163560.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:55:17,237 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163567.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:55:26,423 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 01:55:29,002 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.410e+02 2.798e+02 3.588e+02 7.290e+02, threshold=5.595e+02, percent-clipped=1.0 +2023-02-07 01:55:29,220 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163585.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:55:37,723 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 01:55:40,620 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=163602.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:55:43,572 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-07 01:55:45,155 INFO [train.py:901] (1/4) Epoch 21, batch 1950, loss[loss=0.1748, simple_loss=0.2606, pruned_loss=0.04454, over 8079.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2899, pruned_loss=0.06341, over 1614754.40 frames. ], batch size: 21, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:55:58,505 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 01:56:01,711 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-07 01:56:21,632 INFO [train.py:901] (1/4) Epoch 21, batch 2000, loss[loss=0.2165, simple_loss=0.2965, pruned_loss=0.06821, over 8510.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2901, pruned_loss=0.0632, over 1616788.57 frames. ], batch size: 29, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:56:29,908 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8636, 2.0112, 1.9925, 1.6379, 2.0737, 1.6733, 1.2478, 1.8869], + device='cuda:1'), covar=tensor([0.0458, 0.0270, 0.0210, 0.0430, 0.0274, 0.0608, 0.0622, 0.0229], + device='cuda:1'), in_proj_covar=tensor([0.0446, 0.0384, 0.0336, 0.0438, 0.0367, 0.0528, 0.0387, 0.0414], + device='cuda:1'), out_proj_covar=tensor([1.2002e-04, 1.0088e-04, 8.8661e-05, 1.1577e-04, 9.6821e-05, 1.4969e-04, + 1.0456e-04, 1.1004e-04], device='cuda:1') +2023-02-07 01:56:39,053 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.546e+02 3.013e+02 3.975e+02 6.874e+02, threshold=6.025e+02, percent-clipped=4.0 +2023-02-07 01:56:55,198 INFO [train.py:901] (1/4) Epoch 21, batch 2050, loss[loss=0.1712, simple_loss=0.2508, pruned_loss=0.04579, over 7967.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2896, pruned_loss=0.06321, over 1613843.97 frames. ], batch size: 21, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:57:00,606 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163717.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:57:05,684 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.4576, 1.7330, 5.5878, 2.3128, 5.0433, 4.7536, 5.1682, 5.0427], + device='cuda:1'), covar=tensor([0.0563, 0.4912, 0.0419, 0.3784, 0.0961, 0.0890, 0.0529, 0.0523], + device='cuda:1'), in_proj_covar=tensor([0.0624, 0.0640, 0.0685, 0.0622, 0.0701, 0.0603, 0.0602, 0.0670], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:57:30,310 INFO [train.py:901] (1/4) Epoch 21, batch 2100, loss[loss=0.2163, simple_loss=0.2987, pruned_loss=0.06699, over 8245.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2895, pruned_loss=0.06337, over 1609065.51 frames. ], batch size: 24, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:57:43,622 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 +2023-02-07 01:57:48,642 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.575e+02 2.946e+02 3.630e+02 8.805e+02, threshold=5.893e+02, percent-clipped=3.0 +2023-02-07 01:58:04,870 INFO [train.py:901] (1/4) Epoch 21, batch 2150, loss[loss=0.2358, simple_loss=0.3231, pruned_loss=0.07426, over 8361.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2889, pruned_loss=0.0629, over 1612766.23 frames. ], batch size: 24, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:58:14,673 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163823.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:58:31,298 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163848.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:58:39,335 INFO [train.py:901] (1/4) Epoch 21, batch 2200, loss[loss=0.2111, simple_loss=0.2918, pruned_loss=0.06516, over 8248.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2878, pruned_loss=0.06231, over 1608332.20 frames. ], batch size: 22, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:58:58,248 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.475e+02 2.987e+02 3.670e+02 7.762e+02, threshold=5.973e+02, percent-clipped=3.0 +2023-02-07 01:59:15,125 INFO [train.py:901] (1/4) Epoch 21, batch 2250, loss[loss=0.2378, simple_loss=0.3112, pruned_loss=0.08214, over 6807.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2887, pruned_loss=0.06254, over 1611883.67 frames. ], batch size: 72, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 01:59:21,984 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2137, 4.2404, 3.7929, 2.0563, 3.7540, 3.7182, 3.7525, 3.6232], + device='cuda:1'), covar=tensor([0.0791, 0.0536, 0.1081, 0.4514, 0.0926, 0.1001, 0.1238, 0.0965], + device='cuda:1'), in_proj_covar=tensor([0.0519, 0.0433, 0.0435, 0.0538, 0.0426, 0.0441, 0.0421, 0.0382], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:59:30,188 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.4928, 5.6367, 4.8602, 2.5951, 4.9358, 5.2925, 5.1674, 5.0844], + device='cuda:1'), covar=tensor([0.0583, 0.0406, 0.0949, 0.4683, 0.0839, 0.0788, 0.1006, 0.0701], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0433, 0.0435, 0.0538, 0.0427, 0.0441, 0.0421, 0.0383], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 01:59:33,087 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6922, 1.9582, 2.1053, 1.4031, 2.2164, 1.5808, 0.6697, 1.9293], + device='cuda:1'), covar=tensor([0.0631, 0.0386, 0.0297, 0.0588, 0.0361, 0.0843, 0.0962, 0.0305], + device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0384, 0.0336, 0.0439, 0.0369, 0.0526, 0.0387, 0.0413], + device='cuda:1'), out_proj_covar=tensor([1.2041e-04, 1.0092e-04, 8.8536e-05, 1.1599e-04, 9.7212e-05, 1.4906e-04, + 1.0461e-04, 1.0979e-04], device='cuda:1') +2023-02-07 01:59:49,141 INFO [train.py:901] (1/4) Epoch 21, batch 2300, loss[loss=0.1984, simple_loss=0.2777, pruned_loss=0.05956, over 8101.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2883, pruned_loss=0.06223, over 1614689.47 frames. ], batch size: 23, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 01:59:58,900 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163973.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:00:08,071 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 2.361e+02 2.889e+02 3.736e+02 8.411e+02, threshold=5.778e+02, percent-clipped=4.0 +2023-02-07 02:00:17,861 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163998.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:00:20,568 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-07 02:00:26,181 INFO [train.py:901] (1/4) Epoch 21, batch 2350, loss[loss=0.1756, simple_loss=0.2678, pruned_loss=0.04174, over 8235.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2886, pruned_loss=0.06183, over 1616232.11 frames. ], batch size: 22, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 02:00:39,166 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 02:00:46,898 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1701, 1.3264, 1.3114, 1.0170, 1.3719, 1.1057, 0.3278, 1.2784], + device='cuda:1'), covar=tensor([0.0377, 0.0273, 0.0213, 0.0375, 0.0290, 0.0560, 0.0683, 0.0190], + device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0383, 0.0334, 0.0437, 0.0368, 0.0525, 0.0386, 0.0412], + device='cuda:1'), out_proj_covar=tensor([1.2028e-04, 1.0044e-04, 8.8028e-05, 1.1556e-04, 9.7141e-05, 1.4874e-04, + 1.0434e-04, 1.0937e-04], device='cuda:1') +2023-02-07 02:01:01,227 INFO [train.py:901] (1/4) Epoch 21, batch 2400, loss[loss=0.1998, simple_loss=0.2599, pruned_loss=0.06987, over 7533.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2893, pruned_loss=0.0626, over 1616302.88 frames. ], batch size: 18, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 02:01:19,285 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.419e+02 2.926e+02 3.800e+02 6.132e+02, threshold=5.852e+02, percent-clipped=4.0 +2023-02-07 02:01:37,288 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164108.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:01:37,736 INFO [train.py:901] (1/4) Epoch 21, batch 2450, loss[loss=0.2822, simple_loss=0.3453, pruned_loss=0.1095, over 8288.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2889, pruned_loss=0.06262, over 1607638.90 frames. ], batch size: 23, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 02:01:54,109 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 02:02:12,735 INFO [train.py:901] (1/4) Epoch 21, batch 2500, loss[loss=0.2176, simple_loss=0.3, pruned_loss=0.06759, over 8469.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2881, pruned_loss=0.06227, over 1610540.50 frames. ], batch size: 25, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:02:22,147 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164173.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:02:30,792 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.421e+02 3.174e+02 4.025e+02 1.090e+03, threshold=6.349e+02, percent-clipped=9.0 +2023-02-07 02:02:46,233 INFO [train.py:901] (1/4) Epoch 21, batch 2550, loss[loss=0.2259, simple_loss=0.3105, pruned_loss=0.07062, over 8456.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2894, pruned_loss=0.06323, over 1613157.66 frames. ], batch size: 29, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:02:55,028 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1695, 2.2332, 1.8348, 2.8618, 1.3221, 1.6917, 2.0027, 2.2569], + device='cuda:1'), covar=tensor([0.0667, 0.0803, 0.0900, 0.0362, 0.1189, 0.1304, 0.0947, 0.0835], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0195, 0.0244, 0.0212, 0.0204, 0.0244, 0.0250, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 02:03:22,649 INFO [train.py:901] (1/4) Epoch 21, batch 2600, loss[loss=0.1608, simple_loss=0.2421, pruned_loss=0.03978, over 7797.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.289, pruned_loss=0.06299, over 1612432.74 frames. ], batch size: 19, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:03:40,882 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.272e+02 2.670e+02 3.622e+02 6.852e+02, threshold=5.341e+02, percent-clipped=1.0 +2023-02-07 02:03:56,835 INFO [train.py:901] (1/4) Epoch 21, batch 2650, loss[loss=0.1991, simple_loss=0.2864, pruned_loss=0.05593, over 8030.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2886, pruned_loss=0.06273, over 1612327.06 frames. ], batch size: 22, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:04:33,159 INFO [train.py:901] (1/4) Epoch 21, batch 2700, loss[loss=0.2687, simple_loss=0.3341, pruned_loss=0.1017, over 7023.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2878, pruned_loss=0.06205, over 1610682.70 frames. ], batch size: 71, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:04:46,946 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164378.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:04:52,076 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.228e+02 2.697e+02 3.361e+02 7.045e+02, threshold=5.394e+02, percent-clipped=4.0 +2023-02-07 02:05:07,796 INFO [train.py:901] (1/4) Epoch 21, batch 2750, loss[loss=0.2386, simple_loss=0.319, pruned_loss=0.07908, over 8331.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2882, pruned_loss=0.06232, over 1608815.14 frames. ], batch size: 26, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:05:36,818 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=164452.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:05:42,225 INFO [train.py:901] (1/4) Epoch 21, batch 2800, loss[loss=0.1638, simple_loss=0.2531, pruned_loss=0.03731, over 7928.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2879, pruned_loss=0.06219, over 1610576.26 frames. ], batch size: 20, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:06:02,577 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.305e+02 2.813e+02 3.760e+02 7.507e+02, threshold=5.625e+02, percent-clipped=3.0 +2023-02-07 02:06:18,042 INFO [train.py:901] (1/4) Epoch 21, batch 2850, loss[loss=0.1698, simple_loss=0.2567, pruned_loss=0.04149, over 8024.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2887, pruned_loss=0.06325, over 1604846.86 frames. ], batch size: 22, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:06:23,425 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=164517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:06:51,355 INFO [train.py:901] (1/4) Epoch 21, batch 2900, loss[loss=0.2812, simple_loss=0.364, pruned_loss=0.09921, over 8294.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2905, pruned_loss=0.06386, over 1606183.63 frames. ], batch size: 23, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:06:56,992 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164567.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:07:09,738 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 02:07:11,676 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.734e+02 2.599e+02 3.265e+02 4.069e+02 1.074e+03, threshold=6.531e+02, percent-clipped=8.0 +2023-02-07 02:07:11,869 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:07:19,714 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164596.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:07:28,723 INFO [train.py:901] (1/4) Epoch 21, batch 2950, loss[loss=0.2171, simple_loss=0.3063, pruned_loss=0.0639, over 8345.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2915, pruned_loss=0.06397, over 1612596.51 frames. ], batch size: 24, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:07:44,450 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164632.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:08:02,292 INFO [train.py:901] (1/4) Epoch 21, batch 3000, loss[loss=0.1876, simple_loss=0.2726, pruned_loss=0.05127, over 8291.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2918, pruned_loss=0.0639, over 1615027.42 frames. ], batch size: 23, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:08:02,293 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 02:08:15,069 INFO [train.py:935] (1/4) Epoch 21, validation: loss=0.1742, simple_loss=0.2744, pruned_loss=0.03706, over 944034.00 frames. +2023-02-07 02:08:15,070 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 02:08:26,767 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164676.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:08:33,560 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.728e+02 2.380e+02 2.886e+02 3.399e+02 6.002e+02, threshold=5.772e+02, percent-clipped=0.0 +2023-02-07 02:08:46,925 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-07 02:08:49,848 INFO [train.py:901] (1/4) Epoch 21, batch 3050, loss[loss=0.1935, simple_loss=0.268, pruned_loss=0.0595, over 7444.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2918, pruned_loss=0.06433, over 1614827.24 frames. ], batch size: 17, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:08:59,339 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=164722.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:09:08,319 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6771, 5.8706, 5.0045, 2.5629, 5.1006, 5.6039, 5.4923, 5.3792], + device='cuda:1'), covar=tensor([0.0608, 0.0436, 0.1010, 0.4548, 0.0812, 0.0776, 0.1025, 0.0599], + device='cuda:1'), in_proj_covar=tensor([0.0512, 0.0428, 0.0428, 0.0529, 0.0420, 0.0432, 0.0411, 0.0377], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:09:25,492 INFO [train.py:901] (1/4) Epoch 21, batch 3100, loss[loss=0.188, simple_loss=0.2693, pruned_loss=0.05339, over 7932.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2912, pruned_loss=0.0642, over 1615195.67 frames. ], batch size: 20, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:09:29,031 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164764.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:09:38,547 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8127, 2.3979, 1.9365, 2.1677, 2.1467, 1.8385, 2.0776, 2.1498], + device='cuda:1'), covar=tensor([0.0994, 0.0356, 0.0813, 0.0438, 0.0542, 0.1043, 0.0709, 0.0694], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0234, 0.0333, 0.0306, 0.0296, 0.0331, 0.0341, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 02:09:39,120 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2128, 4.1848, 3.7762, 1.9858, 3.6875, 3.7754, 3.7126, 3.6130], + device='cuda:1'), covar=tensor([0.0832, 0.0628, 0.1208, 0.4705, 0.1018, 0.0970, 0.1389, 0.0931], + device='cuda:1'), in_proj_covar=tensor([0.0509, 0.0426, 0.0426, 0.0526, 0.0417, 0.0430, 0.0409, 0.0376], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:09:43,641 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.375e+02 2.980e+02 3.572e+02 8.800e+02, threshold=5.960e+02, percent-clipped=5.0 +2023-02-07 02:09:59,127 INFO [train.py:901] (1/4) Epoch 21, batch 3150, loss[loss=0.2219, simple_loss=0.3067, pruned_loss=0.0685, over 8244.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2909, pruned_loss=0.06425, over 1614200.27 frames. ], batch size: 24, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:10:08,647 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164823.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:10:12,645 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9600, 1.5519, 3.3909, 1.5628, 2.4964, 3.7752, 3.8642, 3.2363], + device='cuda:1'), covar=tensor([0.1143, 0.1846, 0.0351, 0.2166, 0.0943, 0.0217, 0.0366, 0.0554], + device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0320, 0.0288, 0.0314, 0.0304, 0.0261, 0.0410, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 02:10:19,451 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164837.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:10:26,877 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164848.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:10:34,953 INFO [train.py:901] (1/4) Epoch 21, batch 3200, loss[loss=0.1807, simple_loss=0.2651, pruned_loss=0.04813, over 7799.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2903, pruned_loss=0.06415, over 1610469.48 frames. ], batch size: 20, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:10:54,110 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.324e+02 2.650e+02 3.384e+02 7.808e+02, threshold=5.299e+02, percent-clipped=1.0 +2023-02-07 02:10:55,685 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164888.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:11:09,475 INFO [train.py:901] (1/4) Epoch 21, batch 3250, loss[loss=0.1639, simple_loss=0.241, pruned_loss=0.04338, over 7664.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2905, pruned_loss=0.06393, over 1611863.57 frames. ], batch size: 19, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:11:12,444 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164913.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:11:23,942 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=164930.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:11:30,775 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=164940.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:11:44,786 INFO [train.py:901] (1/4) Epoch 21, batch 3300, loss[loss=0.2117, simple_loss=0.2983, pruned_loss=0.06255, over 8641.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2903, pruned_loss=0.0637, over 1617789.28 frames. ], batch size: 39, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:12:05,213 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.295e+02 2.742e+02 3.217e+02 7.829e+02, threshold=5.483e+02, percent-clipped=4.0 +2023-02-07 02:12:20,626 INFO [train.py:901] (1/4) Epoch 21, batch 3350, loss[loss=0.211, simple_loss=0.2973, pruned_loss=0.06235, over 8343.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2916, pruned_loss=0.06451, over 1613156.27 frames. ], batch size: 24, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:12:28,061 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=165020.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:12:45,357 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165045.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:12:52,356 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165055.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:12:54,954 INFO [train.py:901] (1/4) Epoch 21, batch 3400, loss[loss=0.1943, simple_loss=0.2749, pruned_loss=0.05686, over 8664.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2911, pruned_loss=0.06407, over 1614255.90 frames. ], batch size: 39, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:12:59,414 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.64 vs. limit=5.0 +2023-02-07 02:13:15,620 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 2.300e+02 2.821e+02 3.884e+02 1.046e+03, threshold=5.643e+02, percent-clipped=8.0 +2023-02-07 02:13:20,661 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165093.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:13:31,317 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=165108.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:13:31,884 INFO [train.py:901] (1/4) Epoch 21, batch 3450, loss[loss=0.3009, simple_loss=0.364, pruned_loss=0.1189, over 7226.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2902, pruned_loss=0.06348, over 1613510.29 frames. ], batch size: 71, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:13:38,107 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:13:49,300 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165135.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:13:53,299 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165141.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:14:05,183 INFO [train.py:901] (1/4) Epoch 21, batch 3500, loss[loss=0.1679, simple_loss=0.2535, pruned_loss=0.04118, over 7200.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2905, pruned_loss=0.06358, over 1616640.86 frames. ], batch size: 16, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:14:07,492 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.34 vs. limit=5.0 +2023-02-07 02:14:10,616 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 02:14:19,678 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7648, 1.6208, 2.5344, 1.6757, 1.3048, 2.4035, 0.5161, 1.5186], + device='cuda:1'), covar=tensor([0.1754, 0.1721, 0.0341, 0.1418, 0.2929, 0.0533, 0.2301, 0.1572], + device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0196, 0.0128, 0.0222, 0.0272, 0.0135, 0.0170, 0.0191], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 02:14:24,648 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.436e+02 2.745e+02 3.695e+02 8.606e+02, threshold=5.490e+02, percent-clipped=3.0 +2023-02-07 02:14:31,313 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-07 02:14:41,283 INFO [train.py:901] (1/4) Epoch 21, batch 3550, loss[loss=0.2237, simple_loss=0.3067, pruned_loss=0.07034, over 8252.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2894, pruned_loss=0.06314, over 1615901.15 frames. ], batch size: 22, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:14:51,651 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165223.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:15:15,595 INFO [train.py:901] (1/4) Epoch 21, batch 3600, loss[loss=0.2087, simple_loss=0.2962, pruned_loss=0.06056, over 8324.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2907, pruned_loss=0.06399, over 1614565.03 frames. ], batch size: 26, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:15:34,165 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.347e+02 2.942e+02 3.699e+02 7.087e+02, threshold=5.884e+02, percent-clipped=2.0 +2023-02-07 02:15:36,935 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8515, 1.4788, 1.6329, 1.3274, 0.9132, 1.4290, 1.6020, 1.3764], + device='cuda:1'), covar=tensor([0.0533, 0.1283, 0.1683, 0.1444, 0.0633, 0.1517, 0.0715, 0.0694], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0189, 0.0158, 0.0099, 0.0161, 0.0112, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 02:15:44,469 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165301.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:15:51,143 INFO [train.py:901] (1/4) Epoch 21, batch 3650, loss[loss=0.1954, simple_loss=0.2779, pruned_loss=0.05642, over 8519.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2896, pruned_loss=0.06362, over 1613506.34 frames. ], batch size: 29, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:15:52,693 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165311.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:15:53,562 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 02:15:56,696 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3814, 2.7995, 2.3121, 3.7435, 1.6294, 2.1911, 2.1928, 2.6974], + device='cuda:1'), covar=tensor([0.0699, 0.0751, 0.0820, 0.0305, 0.1137, 0.1135, 0.1035, 0.0773], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0196, 0.0244, 0.0212, 0.0205, 0.0246, 0.0250, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 02:16:03,316 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165326.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:16:08,101 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6144, 2.3669, 4.2219, 1.4170, 3.0507, 2.2967, 1.8812, 3.0525], + device='cuda:1'), covar=tensor([0.2041, 0.2678, 0.0869, 0.4747, 0.1871, 0.3217, 0.2343, 0.2370], + device='cuda:1'), in_proj_covar=tensor([0.0523, 0.0596, 0.0557, 0.0639, 0.0645, 0.0592, 0.0535, 0.0634], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:16:10,832 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165336.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:16:16,763 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 02:16:25,985 INFO [train.py:901] (1/4) Epoch 21, batch 3700, loss[loss=0.324, simple_loss=0.3759, pruned_loss=0.136, over 6827.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2901, pruned_loss=0.06412, over 1614402.61 frames. ], batch size: 72, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:16:44,015 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.608e+02 2.402e+02 2.885e+02 3.854e+02 8.848e+02, threshold=5.771e+02, percent-clipped=5.0 +2023-02-07 02:16:47,623 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165391.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:16:59,580 INFO [train.py:901] (1/4) Epoch 21, batch 3750, loss[loss=0.2388, simple_loss=0.3238, pruned_loss=0.07688, over 8570.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2903, pruned_loss=0.06429, over 1615041.84 frames. ], batch size: 39, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:17:04,495 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165416.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:17:16,250 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1039, 2.3045, 1.9330, 2.8554, 1.3644, 1.7509, 1.8956, 2.2563], + device='cuda:1'), covar=tensor([0.0714, 0.0734, 0.0878, 0.0348, 0.1134, 0.1270, 0.0923, 0.0832], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0196, 0.0244, 0.0212, 0.0205, 0.0246, 0.0251, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 02:17:36,645 INFO [train.py:901] (1/4) Epoch 21, batch 3800, loss[loss=0.2113, simple_loss=0.2951, pruned_loss=0.06382, over 8239.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2905, pruned_loss=0.0644, over 1617584.68 frames. ], batch size: 22, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:17:49,666 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6706, 1.4750, 1.8037, 1.2734, 0.9125, 1.5480, 1.6052, 1.5642], + device='cuda:1'), covar=tensor([0.0536, 0.1236, 0.1547, 0.1486, 0.0555, 0.1413, 0.0635, 0.0599], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0160, 0.0099, 0.0162, 0.0113, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 02:17:50,392 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165479.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:17:54,320 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=165485.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:17:54,906 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.401e+02 2.925e+02 3.673e+02 6.793e+02, threshold=5.851e+02, percent-clipped=2.0 +2023-02-07 02:17:59,010 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7052, 1.5247, 3.1927, 1.3354, 2.3937, 3.4320, 3.5924, 2.9474], + device='cuda:1'), covar=tensor([0.1206, 0.1686, 0.0344, 0.2149, 0.0924, 0.0255, 0.0568, 0.0577], + device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0321, 0.0287, 0.0315, 0.0308, 0.0263, 0.0412, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 02:18:07,150 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165504.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:18:10,365 INFO [train.py:901] (1/4) Epoch 21, batch 3850, loss[loss=0.1798, simple_loss=0.2621, pruned_loss=0.04877, over 5915.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2908, pruned_loss=0.06447, over 1613673.42 frames. ], batch size: 13, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:18:18,546 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 02:18:31,966 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3063, 2.1498, 1.6842, 1.8998, 1.8412, 1.4541, 1.7498, 1.6519], + device='cuda:1'), covar=tensor([0.1268, 0.0456, 0.1239, 0.0529, 0.0657, 0.1505, 0.0839, 0.0801], + device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0235, 0.0334, 0.0310, 0.0298, 0.0336, 0.0345, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 02:18:46,117 INFO [train.py:901] (1/4) Epoch 21, batch 3900, loss[loss=0.1988, simple_loss=0.2829, pruned_loss=0.05736, over 8495.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2904, pruned_loss=0.064, over 1614014.65 frames. ], batch size: 26, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:18:53,038 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165569.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:19:05,130 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.182e+02 2.809e+02 3.459e+02 6.713e+02, threshold=5.619e+02, percent-clipped=4.0 +2023-02-07 02:19:14,928 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165600.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:19:20,732 INFO [train.py:901] (1/4) Epoch 21, batch 3950, loss[loss=0.1946, simple_loss=0.2529, pruned_loss=0.06817, over 7426.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2905, pruned_loss=0.06422, over 1610668.50 frames. ], batch size: 17, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:19:55,187 INFO [train.py:901] (1/4) Epoch 21, batch 4000, loss[loss=0.1961, simple_loss=0.2673, pruned_loss=0.06245, over 7430.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2903, pruned_loss=0.06383, over 1613302.30 frames. ], batch size: 17, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:20:15,749 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 2.370e+02 2.936e+02 3.785e+02 6.204e+02, threshold=5.872e+02, percent-clipped=2.0 +2023-02-07 02:20:31,250 INFO [train.py:901] (1/4) Epoch 21, batch 4050, loss[loss=0.2255, simple_loss=0.3101, pruned_loss=0.07041, over 8245.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2907, pruned_loss=0.06312, over 1615531.42 frames. ], batch size: 24, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:20:47,659 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-07 02:21:03,807 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 +2023-02-07 02:21:04,769 INFO [train.py:901] (1/4) Epoch 21, batch 4100, loss[loss=0.1735, simple_loss=0.2435, pruned_loss=0.05174, over 7216.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2895, pruned_loss=0.06239, over 1617087.03 frames. ], batch size: 16, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:21:11,130 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:21:24,843 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.510e+02 3.105e+02 3.860e+02 6.931e+02, threshold=6.209e+02, percent-clipped=6.0 +2023-02-07 02:21:41,904 INFO [train.py:901] (1/4) Epoch 21, batch 4150, loss[loss=0.2054, simple_loss=0.2873, pruned_loss=0.06177, over 8082.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2895, pruned_loss=0.06271, over 1616592.47 frames. ], batch size: 21, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:21:48,858 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6150, 1.5316, 2.1134, 1.4089, 1.2265, 2.0772, 0.3331, 1.2685], + device='cuda:1'), covar=tensor([0.1642, 0.1634, 0.0416, 0.1014, 0.2811, 0.0425, 0.2158, 0.1306], + device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0195, 0.0128, 0.0221, 0.0272, 0.0135, 0.0171, 0.0191], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 02:21:50,126 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165821.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:22:13,875 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165856.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:22:15,091 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 02:22:15,757 INFO [train.py:901] (1/4) Epoch 21, batch 4200, loss[loss=0.2026, simple_loss=0.2911, pruned_loss=0.05707, over 8292.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2898, pruned_loss=0.06324, over 1614563.53 frames. ], batch size: 23, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:22:30,505 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165881.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:22:33,705 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.320e+02 2.907e+02 3.705e+02 7.802e+02, threshold=5.814e+02, percent-clipped=2.0 +2023-02-07 02:22:37,065 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 02:22:50,802 INFO [train.py:901] (1/4) Epoch 21, batch 4250, loss[loss=0.1659, simple_loss=0.2489, pruned_loss=0.04142, over 7924.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2898, pruned_loss=0.06262, over 1614861.34 frames. ], batch size: 20, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:22:53,703 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=165913.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:23:14,530 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5933, 2.0125, 3.2322, 1.4213, 2.3982, 2.0393, 1.6663, 2.3458], + device='cuda:1'), covar=tensor([0.1914, 0.2553, 0.0805, 0.4473, 0.1764, 0.3193, 0.2342, 0.2315], + device='cuda:1'), in_proj_covar=tensor([0.0523, 0.0597, 0.0555, 0.0639, 0.0645, 0.0594, 0.0536, 0.0635], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:23:26,542 INFO [train.py:901] (1/4) Epoch 21, batch 4300, loss[loss=0.2121, simple_loss=0.2997, pruned_loss=0.06219, over 8353.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2895, pruned_loss=0.0626, over 1618786.35 frames. ], batch size: 24, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:23:44,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.579e+02 2.284e+02 2.728e+02 3.396e+02 7.954e+02, threshold=5.457e+02, percent-clipped=4.0 +2023-02-07 02:23:50,846 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5516, 1.7966, 2.6791, 1.4229, 2.0088, 1.8126, 1.6601, 1.9640], + device='cuda:1'), covar=tensor([0.1814, 0.2439, 0.0823, 0.4246, 0.1767, 0.3159, 0.2192, 0.2176], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0595, 0.0551, 0.0636, 0.0642, 0.0591, 0.0534, 0.0632], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:24:01,653 INFO [train.py:901] (1/4) Epoch 21, batch 4350, loss[loss=0.2001, simple_loss=0.2763, pruned_loss=0.06199, over 7648.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2881, pruned_loss=0.06228, over 1618805.18 frames. ], batch size: 19, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:24:11,729 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 02:24:15,222 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166028.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:24:16,540 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5132, 1.6011, 2.2112, 1.4264, 1.6140, 1.7299, 1.5358, 1.5705], + device='cuda:1'), covar=tensor([0.1850, 0.2575, 0.0818, 0.4182, 0.1777, 0.3302, 0.2253, 0.2002], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0593, 0.0551, 0.0635, 0.0641, 0.0590, 0.0533, 0.0631], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:24:36,835 INFO [train.py:901] (1/4) Epoch 21, batch 4400, loss[loss=0.2333, simple_loss=0.3119, pruned_loss=0.07731, over 8598.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2877, pruned_loss=0.06199, over 1619655.80 frames. ], batch size: 34, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:24:45,739 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166072.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 02:24:54,411 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 02:24:55,055 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.482e+02 3.095e+02 3.863e+02 7.424e+02, threshold=6.191e+02, percent-clipped=10.0 +2023-02-07 02:25:10,665 INFO [train.py:901] (1/4) Epoch 21, batch 4450, loss[loss=0.2344, simple_loss=0.304, pruned_loss=0.08237, over 6697.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2862, pruned_loss=0.06076, over 1617793.63 frames. ], batch size: 71, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:25:12,802 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166112.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:25:22,690 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5362, 2.6966, 3.0163, 1.8819, 3.2827, 2.2221, 1.6203, 2.2166], + device='cuda:1'), covar=tensor([0.0764, 0.0420, 0.0289, 0.0770, 0.0486, 0.0791, 0.0993, 0.0547], + device='cuda:1'), in_proj_covar=tensor([0.0454, 0.0387, 0.0341, 0.0441, 0.0374, 0.0533, 0.0391, 0.0415], + device='cuda:1'), out_proj_covar=tensor([1.2196e-04, 1.0144e-04, 8.9743e-05, 1.1657e-04, 9.8624e-05, 1.5080e-04, + 1.0559e-04, 1.1017e-04], device='cuda:1') +2023-02-07 02:25:27,716 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 02:25:45,424 INFO [train.py:901] (1/4) Epoch 21, batch 4500, loss[loss=0.2213, simple_loss=0.3165, pruned_loss=0.06303, over 8469.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2865, pruned_loss=0.06124, over 1617792.22 frames. ], batch size: 25, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:25:50,222 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 02:25:50,296 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166165.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:25:50,366 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166165.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:26:05,003 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.688e+02 3.191e+02 4.415e+02 1.086e+03, threshold=6.382e+02, percent-clipped=9.0 +2023-02-07 02:26:20,604 INFO [train.py:901] (1/4) Epoch 21, batch 4550, loss[loss=0.2418, simple_loss=0.3178, pruned_loss=0.08284, over 8284.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2866, pruned_loss=0.06175, over 1613367.15 frames. ], batch size: 23, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:26:27,578 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2739, 2.0015, 2.7012, 2.1907, 2.6050, 2.3014, 2.0237, 1.4359], + device='cuda:1'), covar=tensor([0.5160, 0.5107, 0.1918, 0.3804, 0.2647, 0.3031, 0.1924, 0.5277], + device='cuda:1'), in_proj_covar=tensor([0.0935, 0.0972, 0.0794, 0.0932, 0.0989, 0.0885, 0.0741, 0.0820], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 02:26:33,047 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166227.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:26:54,689 INFO [train.py:901] (1/4) Epoch 21, batch 4600, loss[loss=0.1943, simple_loss=0.2831, pruned_loss=0.05269, over 8195.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2863, pruned_loss=0.06158, over 1613627.56 frames. ], batch size: 23, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:27:10,550 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166280.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:27:14,138 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166284.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:27:15,235 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.551e+02 3.310e+02 4.080e+02 7.820e+02, threshold=6.621e+02, percent-clipped=4.0 +2023-02-07 02:27:30,341 INFO [train.py:901] (1/4) Epoch 21, batch 4650, loss[loss=0.2061, simple_loss=0.2911, pruned_loss=0.06057, over 8591.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2882, pruned_loss=0.06255, over 1617394.96 frames. ], batch size: 34, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:27:30,541 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166309.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:27:41,198 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166325.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:28:03,640 INFO [train.py:901] (1/4) Epoch 21, batch 4700, loss[loss=0.226, simple_loss=0.3143, pruned_loss=0.06889, over 8198.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2882, pruned_loss=0.06243, over 1618290.67 frames. ], batch size: 48, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:28:23,899 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.420e+02 2.373e+02 2.801e+02 3.877e+02 1.145e+03, threshold=5.601e+02, percent-clipped=4.0 +2023-02-07 02:28:40,136 INFO [train.py:901] (1/4) Epoch 21, batch 4750, loss[loss=0.1665, simple_loss=0.2461, pruned_loss=0.04348, over 7692.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.289, pruned_loss=0.06262, over 1620793.17 frames. ], batch size: 18, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:28:45,008 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166416.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 02:28:52,966 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 02:28:55,068 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 02:29:14,176 INFO [train.py:901] (1/4) Epoch 21, batch 4800, loss[loss=0.2618, simple_loss=0.3299, pruned_loss=0.09686, over 8332.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2899, pruned_loss=0.06345, over 1619249.68 frames. ], batch size: 26, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:29:30,703 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166483.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:29:33,191 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.407e+02 2.819e+02 3.849e+02 8.316e+02, threshold=5.639e+02, percent-clipped=5.0 +2023-02-07 02:29:39,662 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.98 vs. limit=5.0 +2023-02-07 02:29:42,736 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166499.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:29:43,961 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 02:29:48,835 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166508.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:29:49,314 INFO [train.py:901] (1/4) Epoch 21, batch 4850, loss[loss=0.168, simple_loss=0.2574, pruned_loss=0.03927, over 7799.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2893, pruned_loss=0.06317, over 1619223.98 frames. ], batch size: 20, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:29:49,390 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:03,257 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166527.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:05,913 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166531.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 02:30:09,302 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166536.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:17,926 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166549.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:21,057 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.34 vs. limit=5.0 +2023-02-07 02:30:24,625 INFO [train.py:901] (1/4) Epoch 21, batch 4900, loss[loss=0.1742, simple_loss=0.2579, pruned_loss=0.04522, over 7797.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2898, pruned_loss=0.06357, over 1619142.30 frames. ], batch size: 20, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:30:26,198 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166561.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:43,069 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.429e+02 3.059e+02 4.014e+02 7.599e+02, threshold=6.119e+02, percent-clipped=4.0 +2023-02-07 02:30:54,282 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6212, 1.5199, 2.1441, 1.4068, 1.2965, 2.0596, 0.3090, 1.1947], + device='cuda:1'), covar=tensor([0.1754, 0.1436, 0.0428, 0.1277, 0.2997, 0.0493, 0.2427, 0.1612], + device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0196, 0.0128, 0.0223, 0.0272, 0.0136, 0.0172, 0.0192], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 02:30:58,618 INFO [train.py:901] (1/4) Epoch 21, batch 4950, loss[loss=0.2173, simple_loss=0.2933, pruned_loss=0.07064, over 8464.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.289, pruned_loss=0.06326, over 1616586.58 frames. ], batch size: 25, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:31:09,519 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166624.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:31:34,523 INFO [train.py:901] (1/4) Epoch 21, batch 5000, loss[loss=0.2017, simple_loss=0.2946, pruned_loss=0.05439, over 8465.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2884, pruned_loss=0.06263, over 1620460.96 frames. ], batch size: 29, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:31:41,052 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166669.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:31:52,864 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.664e+02 2.262e+02 2.770e+02 3.475e+02 7.586e+02, threshold=5.540e+02, percent-clipped=2.0 +2023-02-07 02:32:07,639 INFO [train.py:901] (1/4) Epoch 21, batch 5050, loss[loss=0.1663, simple_loss=0.2466, pruned_loss=0.04305, over 7654.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2879, pruned_loss=0.06247, over 1617053.18 frames. ], batch size: 19, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:32:23,026 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 02:32:38,361 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166753.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:32:42,946 INFO [train.py:901] (1/4) Epoch 21, batch 5100, loss[loss=0.1812, simple_loss=0.2594, pruned_loss=0.05149, over 7534.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2886, pruned_loss=0.06302, over 1611552.66 frames. ], batch size: 18, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:32:59,535 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166782.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:33:00,922 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166784.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:33:02,730 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.724e+02 2.499e+02 3.045e+02 3.729e+02 1.083e+03, threshold=6.090e+02, percent-clipped=5.0 +2023-02-07 02:33:02,982 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166787.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 02:33:17,657 INFO [train.py:901] (1/4) Epoch 21, batch 5150, loss[loss=0.2031, simple_loss=0.2864, pruned_loss=0.05994, over 8341.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2891, pruned_loss=0.06309, over 1608911.97 frames. ], batch size: 26, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:33:19,983 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166812.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 02:33:35,606 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9066, 2.2524, 1.8074, 2.8140, 1.3013, 1.5130, 1.8718, 2.2425], + device='cuda:1'), covar=tensor([0.0830, 0.0735, 0.0940, 0.0419, 0.1162, 0.1379, 0.0986, 0.0791], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0194, 0.0241, 0.0210, 0.0203, 0.0241, 0.0249, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 02:33:40,944 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:33:52,915 INFO [train.py:901] (1/4) Epoch 21, batch 5200, loss[loss=0.2103, simple_loss=0.29, pruned_loss=0.06524, over 8345.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2886, pruned_loss=0.06276, over 1608229.66 frames. ], batch size: 24, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:33:54,691 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.00 vs. limit=5.0 +2023-02-07 02:34:01,085 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166871.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:34:08,931 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166880.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:34:13,421 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.417e+02 2.893e+02 3.464e+02 9.071e+02, threshold=5.787e+02, percent-clipped=3.0 +2023-02-07 02:34:17,471 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166893.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:34:22,859 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 02:34:25,856 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166905.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:34:28,364 INFO [train.py:901] (1/4) Epoch 21, batch 5250, loss[loss=0.1903, simple_loss=0.2757, pruned_loss=0.05241, over 8255.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2886, pruned_loss=0.06302, over 1608886.54 frames. ], batch size: 24, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:34:53,837 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166947.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:35:00,977 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166958.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:35:01,476 INFO [train.py:901] (1/4) Epoch 21, batch 5300, loss[loss=0.2094, simple_loss=0.2951, pruned_loss=0.06183, over 8350.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2892, pruned_loss=0.06334, over 1614217.90 frames. ], batch size: 24, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:35:21,123 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166986.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:35:21,589 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.349e+02 2.996e+02 3.802e+02 6.845e+02, threshold=5.992e+02, percent-clipped=3.0 +2023-02-07 02:35:22,371 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4250, 3.8297, 2.4758, 2.9205, 2.9082, 2.3871, 3.1285, 3.1697], + device='cuda:1'), covar=tensor([0.1408, 0.0310, 0.1070, 0.0749, 0.0683, 0.1220, 0.0882, 0.0981], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0235, 0.0335, 0.0308, 0.0298, 0.0334, 0.0344, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 02:35:37,478 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167008.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:35:37,990 INFO [train.py:901] (1/4) Epoch 21, batch 5350, loss[loss=0.164, simple_loss=0.2525, pruned_loss=0.03774, over 7557.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2889, pruned_loss=0.06286, over 1609235.08 frames. ], batch size: 18, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:35:58,871 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167040.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:36:12,088 INFO [train.py:901] (1/4) Epoch 21, batch 5400, loss[loss=0.2276, simple_loss=0.3039, pruned_loss=0.0756, over 8526.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2904, pruned_loss=0.06299, over 1615751.62 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:36:16,470 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167065.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:36:32,170 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.459e+02 3.013e+02 3.547e+02 6.118e+02, threshold=6.026e+02, percent-clipped=1.0 +2023-02-07 02:36:39,934 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167097.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:36:47,948 INFO [train.py:901] (1/4) Epoch 21, batch 5450, loss[loss=0.1941, simple_loss=0.2823, pruned_loss=0.0529, over 8220.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2901, pruned_loss=0.06289, over 1616701.80 frames. ], batch size: 22, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:36:55,798 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:36:58,550 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167122.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:37:01,214 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167126.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:37:12,874 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 02:37:19,298 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.4462, 1.6977, 5.6104, 2.5361, 5.0123, 4.7598, 5.1889, 5.0487], + device='cuda:1'), covar=tensor([0.0505, 0.5096, 0.0348, 0.3567, 0.1038, 0.0845, 0.0494, 0.0534], + device='cuda:1'), in_proj_covar=tensor([0.0630, 0.0643, 0.0695, 0.0629, 0.0707, 0.0605, 0.0605, 0.0677], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:37:24,077 INFO [train.py:901] (1/4) Epoch 21, batch 5500, loss[loss=0.2013, simple_loss=0.2747, pruned_loss=0.06394, over 7654.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.29, pruned_loss=0.06277, over 1616056.52 frames. ], batch size: 19, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:37:43,662 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.538e+02 3.099e+02 3.967e+02 8.838e+02, threshold=6.197e+02, percent-clipped=3.0 +2023-02-07 02:37:58,456 INFO [train.py:901] (1/4) Epoch 21, batch 5550, loss[loss=0.2141, simple_loss=0.2989, pruned_loss=0.06463, over 8512.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2882, pruned_loss=0.06226, over 1612475.63 frames. ], batch size: 39, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:38:01,329 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167212.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:02,582 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:21,139 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:22,456 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167241.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:23,133 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167242.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:34,484 INFO [train.py:901] (1/4) Epoch 21, batch 5600, loss[loss=0.1923, simple_loss=0.2743, pruned_loss=0.05517, over 7529.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2891, pruned_loss=0.06294, over 1615871.71 frames. ], batch size: 18, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:38:34,956 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.31 vs. limit=5.0 +2023-02-07 02:38:36,020 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9861, 1.7986, 6.1296, 2.3082, 5.5141, 5.1611, 5.6883, 5.5353], + device='cuda:1'), covar=tensor([0.0488, 0.4715, 0.0320, 0.3842, 0.0910, 0.0743, 0.0471, 0.0458], + device='cuda:1'), in_proj_covar=tensor([0.0629, 0.0637, 0.0691, 0.0626, 0.0701, 0.0602, 0.0601, 0.0672], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:38:38,123 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167264.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:40,086 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167267.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:54,593 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.492e+02 3.097e+02 3.838e+02 7.086e+02, threshold=6.194e+02, percent-clipped=1.0 +2023-02-07 02:38:54,814 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167289.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:56,076 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167291.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:39:08,032 INFO [train.py:901] (1/4) Epoch 21, batch 5650, loss[loss=0.1994, simple_loss=0.2795, pruned_loss=0.05964, over 7810.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2884, pruned_loss=0.06278, over 1612640.78 frames. ], batch size: 20, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:39:18,773 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 02:39:21,792 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 +2023-02-07 02:39:44,508 INFO [train.py:901] (1/4) Epoch 21, batch 5700, loss[loss=0.2332, simple_loss=0.3176, pruned_loss=0.07445, over 8249.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2887, pruned_loss=0.0632, over 1606958.58 frames. ], batch size: 22, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:40:04,669 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.585e+02 3.206e+02 3.925e+02 8.506e+02, threshold=6.412e+02, percent-clipped=6.0 +2023-02-07 02:40:08,236 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6810, 4.6513, 4.2423, 2.0978, 4.1343, 4.3640, 4.2377, 4.1600], + device='cuda:1'), covar=tensor([0.0651, 0.0454, 0.0991, 0.4417, 0.0842, 0.0873, 0.1188, 0.0727], + device='cuda:1'), in_proj_covar=tensor([0.0523, 0.0430, 0.0430, 0.0534, 0.0425, 0.0440, 0.0420, 0.0382], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:40:16,558 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167406.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:40:17,430 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-07 02:40:18,419 INFO [train.py:901] (1/4) Epoch 21, batch 5750, loss[loss=0.1872, simple_loss=0.2856, pruned_loss=0.04438, over 8749.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2872, pruned_loss=0.06182, over 1606114.62 frames. ], batch size: 30, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:40:24,265 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 02:40:47,691 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167450.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:40:53,630 INFO [train.py:901] (1/4) Epoch 21, batch 5800, loss[loss=0.1481, simple_loss=0.2256, pruned_loss=0.03524, over 7191.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2877, pruned_loss=0.06192, over 1606095.09 frames. ], batch size: 16, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:40:55,798 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167462.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:40:58,561 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7087, 5.8644, 5.0571, 2.5596, 5.0942, 5.4855, 5.3660, 5.2568], + device='cuda:1'), covar=tensor([0.0523, 0.0393, 0.0886, 0.3947, 0.0746, 0.0754, 0.1025, 0.0613], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0427, 0.0427, 0.0530, 0.0422, 0.0436, 0.0416, 0.0380], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:40:59,181 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167466.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:40:59,916 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7566, 5.9438, 5.0946, 2.7668, 5.1253, 5.6559, 5.4273, 5.4036], + device='cuda:1'), covar=tensor([0.0674, 0.0406, 0.1004, 0.4428, 0.0772, 0.0671, 0.1185, 0.0573], + device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0427, 0.0427, 0.0530, 0.0422, 0.0437, 0.0417, 0.0380], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:41:00,728 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167468.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:41:06,728 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3251, 2.7081, 3.0917, 1.4755, 3.1116, 1.9781, 1.6123, 2.2138], + device='cuda:1'), covar=tensor([0.0763, 0.0359, 0.0220, 0.0833, 0.0497, 0.0788, 0.0874, 0.0509], + device='cuda:1'), in_proj_covar=tensor([0.0451, 0.0387, 0.0339, 0.0443, 0.0375, 0.0533, 0.0389, 0.0415], + device='cuda:1'), out_proj_covar=tensor([1.2128e-04, 1.0155e-04, 8.9225e-05, 1.1699e-04, 9.8891e-05, 1.5069e-04, + 1.0509e-04, 1.1005e-04], device='cuda:1') +2023-02-07 02:41:15,047 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.466e+02 2.953e+02 3.603e+02 7.254e+02, threshold=5.907e+02, percent-clipped=1.0 +2023-02-07 02:41:17,997 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167493.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:41:20,749 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167497.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:41:28,628 INFO [train.py:901] (1/4) Epoch 21, batch 5850, loss[loss=0.1716, simple_loss=0.2476, pruned_loss=0.04782, over 7228.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2866, pruned_loss=0.06096, over 1609988.20 frames. ], batch size: 16, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:41:37,829 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167522.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:42:00,639 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7337, 2.1473, 3.2759, 1.5932, 2.4420, 2.1862, 1.8478, 2.4843], + device='cuda:1'), covar=tensor([0.1855, 0.2294, 0.0779, 0.4175, 0.1809, 0.3015, 0.2087, 0.2048], + device='cuda:1'), in_proj_covar=tensor([0.0517, 0.0591, 0.0550, 0.0633, 0.0635, 0.0585, 0.0526, 0.0625], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:42:03,775 INFO [train.py:901] (1/4) Epoch 21, batch 5900, loss[loss=0.2306, simple_loss=0.3256, pruned_loss=0.06782, over 8473.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2858, pruned_loss=0.0607, over 1608640.69 frames. ], batch size: 25, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:42:16,872 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167577.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:42:19,702 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167581.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:42:25,235 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1492, 1.5879, 4.3295, 1.6687, 3.8228, 3.5973, 3.9516, 3.8386], + device='cuda:1'), covar=tensor([0.0622, 0.4166, 0.0572, 0.4052, 0.1239, 0.1030, 0.0635, 0.0656], + device='cuda:1'), in_proj_covar=tensor([0.0631, 0.0641, 0.0695, 0.0632, 0.0706, 0.0607, 0.0608, 0.0681], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:42:25,698 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.257e+02 2.859e+02 3.440e+02 7.059e+02, threshold=5.718e+02, percent-clipped=2.0 +2023-02-07 02:42:34,680 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.99 vs. limit=5.0 +2023-02-07 02:42:40,383 INFO [train.py:901] (1/4) Epoch 21, batch 5950, loss[loss=0.2656, simple_loss=0.3412, pruned_loss=0.09501, over 8574.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2858, pruned_loss=0.0615, over 1601196.14 frames. ], batch size: 34, lr: 3.58e-03, grad_scale: 4.0 +2023-02-07 02:43:14,061 INFO [train.py:901] (1/4) Epoch 21, batch 6000, loss[loss=0.1684, simple_loss=0.2437, pruned_loss=0.0465, over 7710.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2872, pruned_loss=0.06238, over 1607298.60 frames. ], batch size: 18, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:43:14,061 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 02:43:26,396 INFO [train.py:935] (1/4) Epoch 21, validation: loss=0.174, simple_loss=0.2741, pruned_loss=0.03692, over 944034.00 frames. +2023-02-07 02:43:26,397 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 02:43:28,714 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167662.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:43:45,660 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167687.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:43:47,403 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.749e+02 2.382e+02 2.918e+02 3.609e+02 5.587e+02, threshold=5.837e+02, percent-clipped=0.0 +2023-02-07 02:44:01,967 INFO [train.py:901] (1/4) Epoch 21, batch 6050, loss[loss=0.222, simple_loss=0.2992, pruned_loss=0.0724, over 8026.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.288, pruned_loss=0.06258, over 1610855.85 frames. ], batch size: 22, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:44:06,715 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-07 02:44:22,066 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167737.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:44:38,028 INFO [train.py:901] (1/4) Epoch 21, batch 6100, loss[loss=0.2203, simple_loss=0.306, pruned_loss=0.0673, over 8467.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2879, pruned_loss=0.06218, over 1612636.10 frames. ], batch size: 29, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:44:56,062 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167785.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:44:57,220 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 02:44:58,532 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.708e+02 2.390e+02 3.045e+02 3.849e+02 6.701e+02, threshold=6.089e+02, percent-clipped=2.0 +2023-02-07 02:45:02,108 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167794.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:45:13,115 INFO [train.py:901] (1/4) Epoch 21, batch 6150, loss[loss=0.1671, simple_loss=0.2525, pruned_loss=0.04081, over 7663.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2881, pruned_loss=0.06273, over 1610759.39 frames. ], batch size: 19, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:45:30,428 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167833.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:45:33,106 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167837.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:45:48,358 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167858.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:45:48,818 INFO [train.py:901] (1/4) Epoch 21, batch 6200, loss[loss=0.1693, simple_loss=0.2509, pruned_loss=0.04383, over 7454.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2871, pruned_loss=0.06219, over 1608800.83 frames. ], batch size: 17, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:45:51,052 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167862.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:46:09,469 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.266e+02 2.776e+02 3.727e+02 8.167e+02, threshold=5.552e+02, percent-clipped=4.0 +2023-02-07 02:46:16,723 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 02:46:23,395 INFO [train.py:901] (1/4) Epoch 21, batch 6250, loss[loss=0.2057, simple_loss=0.2902, pruned_loss=0.06064, over 8670.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2859, pruned_loss=0.06175, over 1605804.05 frames. ], batch size: 34, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:46:23,602 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167909.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:46:57,214 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.35 vs. limit=5.0 +2023-02-07 02:46:58,910 INFO [train.py:901] (1/4) Epoch 21, batch 6300, loss[loss=0.2016, simple_loss=0.2825, pruned_loss=0.06033, over 8091.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2861, pruned_loss=0.06195, over 1609735.40 frames. ], batch size: 21, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:47:02,546 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 02:47:10,142 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.19 vs. limit=5.0 +2023-02-07 02:47:20,725 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.475e+02 2.869e+02 3.545e+02 9.430e+02, threshold=5.737e+02, percent-clipped=7.0 +2023-02-07 02:47:35,234 INFO [train.py:901] (1/4) Epoch 21, batch 6350, loss[loss=0.1761, simple_loss=0.2623, pruned_loss=0.04497, over 5938.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2859, pruned_loss=0.06164, over 1609316.08 frames. ], batch size: 13, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:48:06,026 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168053.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:48:09,825 INFO [train.py:901] (1/4) Epoch 21, batch 6400, loss[loss=0.1416, simple_loss=0.2293, pruned_loss=0.02698, over 7818.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2861, pruned_loss=0.06174, over 1609884.68 frames. ], batch size: 20, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:48:25,122 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=168081.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:48:30,484 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.236e+02 2.639e+02 3.603e+02 6.999e+02, threshold=5.279e+02, percent-clipped=2.0 +2023-02-07 02:48:45,442 INFO [train.py:901] (1/4) Epoch 21, batch 6450, loss[loss=0.2085, simple_loss=0.2807, pruned_loss=0.0682, over 7791.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2864, pruned_loss=0.06172, over 1612060.34 frames. ], batch size: 19, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:48:59,393 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=168129.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:49:19,836 INFO [train.py:901] (1/4) Epoch 21, batch 6500, loss[loss=0.2188, simple_loss=0.3077, pruned_loss=0.06494, over 8107.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2866, pruned_loss=0.0617, over 1611793.81 frames. ], batch size: 23, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:49:24,789 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168165.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:49:41,359 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.960e+02 2.483e+02 3.129e+02 4.081e+02 1.148e+03, threshold=6.258e+02, percent-clipped=13.0 +2023-02-07 02:49:42,168 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168190.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:49:46,104 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168196.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:49:54,734 INFO [train.py:901] (1/4) Epoch 21, batch 6550, loss[loss=0.1817, simple_loss=0.2652, pruned_loss=0.04917, over 7929.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2875, pruned_loss=0.06199, over 1613223.60 frames. ], batch size: 20, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:50:19,977 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168244.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:50:20,487 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 02:50:24,650 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168251.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:50:29,935 INFO [train.py:901] (1/4) Epoch 21, batch 6600, loss[loss=0.2146, simple_loss=0.3047, pruned_loss=0.06227, over 8513.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2879, pruned_loss=0.06195, over 1613916.44 frames. ], batch size: 29, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:50:32,319 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 02:50:38,741 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 02:50:50,803 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.414e+02 2.830e+02 3.481e+02 7.637e+02, threshold=5.659e+02, percent-clipped=3.0 +2023-02-07 02:51:05,095 INFO [train.py:901] (1/4) Epoch 21, batch 6650, loss[loss=0.1718, simple_loss=0.246, pruned_loss=0.04884, over 7432.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2876, pruned_loss=0.06208, over 1613584.28 frames. ], batch size: 17, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:51:40,097 INFO [train.py:901] (1/4) Epoch 21, batch 6700, loss[loss=0.1961, simple_loss=0.2625, pruned_loss=0.0649, over 7205.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2878, pruned_loss=0.06225, over 1611718.84 frames. ], batch size: 16, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:52:00,447 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.306e+02 2.933e+02 3.476e+02 6.537e+02, threshold=5.866e+02, percent-clipped=2.0 +2023-02-07 02:52:04,742 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5617, 1.8936, 3.0743, 1.4006, 2.1828, 1.9997, 1.6007, 2.1667], + device='cuda:1'), covar=tensor([0.1929, 0.2633, 0.0890, 0.4701, 0.2000, 0.3208, 0.2339, 0.2629], + device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0600, 0.0556, 0.0641, 0.0644, 0.0591, 0.0532, 0.0632], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:52:05,991 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=168397.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:52:14,913 INFO [train.py:901] (1/4) Epoch 21, batch 6750, loss[loss=0.2208, simple_loss=0.2919, pruned_loss=0.0748, over 7916.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2869, pruned_loss=0.06185, over 1610027.74 frames. ], batch size: 20, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:52:19,508 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-07 02:52:45,398 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168452.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:52:50,751 INFO [train.py:901] (1/4) Epoch 21, batch 6800, loss[loss=0.2159, simple_loss=0.3, pruned_loss=0.06593, over 8421.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2868, pruned_loss=0.06172, over 1611644.35 frames. ], batch size: 49, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:52:58,512 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 02:53:04,348 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168477.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:53:12,368 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.333e+02 2.834e+02 3.373e+02 7.883e+02, threshold=5.669e+02, percent-clipped=5.0 +2023-02-07 02:53:20,352 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168500.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:53:26,307 INFO [train.py:901] (1/4) Epoch 21, batch 6850, loss[loss=0.209, simple_loss=0.2922, pruned_loss=0.06287, over 8252.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2873, pruned_loss=0.06216, over 1609673.19 frames. ], batch size: 22, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:53:28,542 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168512.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:53:37,581 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168525.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:53:45,926 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 02:53:54,104 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8875, 1.3062, 1.5690, 1.2926, 1.0129, 1.4675, 1.8186, 1.5008], + device='cuda:1'), covar=tensor([0.0548, 0.1291, 0.1704, 0.1541, 0.0618, 0.1530, 0.0665, 0.0697], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0189, 0.0158, 0.0099, 0.0162, 0.0112, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 02:54:00,744 INFO [train.py:901] (1/4) Epoch 21, batch 6900, loss[loss=0.2055, simple_loss=0.2942, pruned_loss=0.05844, over 8103.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2881, pruned_loss=0.06295, over 1611453.78 frames. ], batch size: 23, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:54:22,253 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.460e+02 2.867e+02 3.613e+02 6.820e+02, threshold=5.733e+02, percent-clipped=1.0 +2023-02-07 02:54:26,393 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=168595.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:54:35,740 INFO [train.py:901] (1/4) Epoch 21, batch 6950, loss[loss=0.2404, simple_loss=0.32, pruned_loss=0.08043, over 8493.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2892, pruned_loss=0.06336, over 1611645.28 frames. ], batch size: 29, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:54:35,962 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8166, 1.7480, 2.5151, 1.6435, 1.3813, 2.3826, 0.6845, 1.6017], + device='cuda:1'), covar=tensor([0.1646, 0.1218, 0.0311, 0.1276, 0.2707, 0.0493, 0.2272, 0.1321], + device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0194, 0.0126, 0.0221, 0.0269, 0.0135, 0.0170, 0.0191], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 02:54:53,419 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 02:55:10,646 INFO [train.py:901] (1/4) Epoch 21, batch 7000, loss[loss=0.2303, simple_loss=0.3086, pruned_loss=0.07604, over 8252.00 frames. ], tot_loss[loss=0.208, simple_loss=0.289, pruned_loss=0.06347, over 1608042.55 frames. ], batch size: 24, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:55:23,595 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8559, 1.3797, 3.9685, 1.4388, 3.5407, 3.3241, 3.6448, 3.5241], + device='cuda:1'), covar=tensor([0.0628, 0.4543, 0.0658, 0.4410, 0.1181, 0.1053, 0.0656, 0.0782], + device='cuda:1'), in_proj_covar=tensor([0.0634, 0.0646, 0.0701, 0.0637, 0.0716, 0.0612, 0.0616, 0.0685], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:55:31,362 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.444e+02 3.041e+02 3.968e+02 8.528e+02, threshold=6.083e+02, percent-clipped=8.0 +2023-02-07 02:55:41,046 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7926, 6.0520, 5.2107, 2.6301, 5.3503, 5.6342, 5.3609, 5.4075], + device='cuda:1'), covar=tensor([0.0523, 0.0329, 0.0764, 0.3765, 0.0724, 0.0773, 0.1064, 0.0570], + device='cuda:1'), in_proj_covar=tensor([0.0523, 0.0431, 0.0430, 0.0534, 0.0423, 0.0440, 0.0419, 0.0383], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:55:45,702 INFO [train.py:901] (1/4) Epoch 21, batch 7050, loss[loss=0.1963, simple_loss=0.287, pruned_loss=0.05276, over 8475.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2887, pruned_loss=0.06341, over 1608807.68 frames. ], batch size: 25, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:55:46,577 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168710.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:55:50,178 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.93 vs. limit=5.0 +2023-02-07 02:56:19,950 INFO [train.py:901] (1/4) Epoch 21, batch 7100, loss[loss=0.2161, simple_loss=0.3013, pruned_loss=0.06547, over 8144.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2866, pruned_loss=0.0622, over 1610403.10 frames. ], batch size: 22, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:56:26,884 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:56:40,684 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.605e+02 3.011e+02 3.811e+02 1.077e+03, threshold=6.022e+02, percent-clipped=4.0 +2023-02-07 02:56:41,628 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5922, 2.1176, 3.2745, 1.7177, 1.6598, 3.1759, 0.9130, 2.1027], + device='cuda:1'), covar=tensor([0.1481, 0.1342, 0.0303, 0.1918, 0.2992, 0.0398, 0.2206, 0.1549], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0196, 0.0127, 0.0222, 0.0271, 0.0136, 0.0172, 0.0192], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 02:56:43,692 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168793.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:56:48,512 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5577, 1.8724, 1.9887, 1.2124, 2.0558, 1.4675, 0.5308, 1.7901], + device='cuda:1'), covar=tensor([0.0638, 0.0382, 0.0284, 0.0607, 0.0437, 0.0932, 0.0865, 0.0307], + device='cuda:1'), in_proj_covar=tensor([0.0453, 0.0391, 0.0342, 0.0443, 0.0375, 0.0534, 0.0389, 0.0417], + device='cuda:1'), out_proj_covar=tensor([1.2173e-04, 1.0258e-04, 9.0082e-05, 1.1684e-04, 9.8664e-05, 1.5105e-04, + 1.0518e-04, 1.1060e-04], device='cuda:1') +2023-02-07 02:56:55,251 INFO [train.py:901] (1/4) Epoch 21, batch 7150, loss[loss=0.1406, simple_loss=0.2275, pruned_loss=0.02689, over 7787.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2868, pruned_loss=0.06238, over 1608702.48 frames. ], batch size: 19, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:57:07,097 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.33 vs. limit=5.0 +2023-02-07 02:57:07,567 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2311, 2.0931, 2.9777, 2.4250, 2.7649, 2.0797, 1.9678, 2.0289], + device='cuda:1'), covar=tensor([0.5211, 0.5199, 0.1769, 0.3469, 0.2418, 0.3916, 0.2514, 0.4214], + device='cuda:1'), in_proj_covar=tensor([0.0939, 0.0973, 0.0800, 0.0941, 0.0997, 0.0890, 0.0746, 0.0820], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 02:57:29,809 INFO [train.py:901] (1/4) Epoch 21, batch 7200, loss[loss=0.1651, simple_loss=0.2562, pruned_loss=0.03705, over 7921.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2877, pruned_loss=0.06266, over 1610852.88 frames. ], batch size: 20, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:57:32,848 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.26 vs. limit=5.0 +2023-02-07 02:57:44,676 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0451, 2.2398, 1.8247, 2.7899, 1.3953, 1.6343, 1.9569, 2.1562], + device='cuda:1'), covar=tensor([0.0770, 0.0832, 0.0977, 0.0391, 0.1111, 0.1366, 0.0945, 0.0835], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0197, 0.0244, 0.0212, 0.0206, 0.0247, 0.0249, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 02:57:51,147 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.788e+02 2.343e+02 3.196e+02 4.097e+02 7.456e+02, threshold=6.392e+02, percent-clipped=6.0 +2023-02-07 02:57:56,891 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 02:58:04,698 INFO [train.py:901] (1/4) Epoch 21, batch 7250, loss[loss=0.1775, simple_loss=0.258, pruned_loss=0.0485, over 7719.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.288, pruned_loss=0.06253, over 1611591.10 frames. ], batch size: 18, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:58:18,037 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5716, 2.4883, 1.8949, 2.2131, 2.1682, 1.5982, 1.9931, 2.0661], + device='cuda:1'), covar=tensor([0.1482, 0.0447, 0.1181, 0.0642, 0.0727, 0.1542, 0.0938, 0.0959], + device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0237, 0.0337, 0.0309, 0.0302, 0.0338, 0.0347, 0.0319], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 02:58:40,068 INFO [train.py:901] (1/4) Epoch 21, batch 7300, loss[loss=0.2203, simple_loss=0.3128, pruned_loss=0.0639, over 8328.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.288, pruned_loss=0.06232, over 1613653.58 frames. ], batch size: 25, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:58:44,984 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168966.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:58:58,201 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168985.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:58:59,432 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168987.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 02:58:59,462 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9756, 3.5274, 1.8276, 2.8222, 2.6229, 1.5286, 2.5930, 2.9940], + device='cuda:1'), covar=tensor([0.1647, 0.0483, 0.1514, 0.0785, 0.0835, 0.2015, 0.1215, 0.0895], + device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0237, 0.0337, 0.0309, 0.0303, 0.0338, 0.0348, 0.0320], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 02:59:00,602 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.375e+02 2.880e+02 4.111e+02 9.346e+02, threshold=5.760e+02, percent-clipped=6.0 +2023-02-07 02:59:02,083 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168991.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:59:14,679 INFO [train.py:901] (1/4) Epoch 21, batch 7350, loss[loss=0.1932, simple_loss=0.2824, pruned_loss=0.05196, over 8623.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2883, pruned_loss=0.0626, over 1616657.43 frames. ], batch size: 49, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:59:35,043 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 02:59:36,091 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 +2023-02-07 02:59:44,847 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5604, 1.9329, 3.3186, 1.4003, 2.4024, 1.9373, 1.6417, 2.4183], + device='cuda:1'), covar=tensor([0.1970, 0.2671, 0.0818, 0.4550, 0.1939, 0.3150, 0.2385, 0.2328], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0603, 0.0558, 0.0644, 0.0645, 0.0596, 0.0534, 0.0632], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 02:59:49,824 INFO [train.py:901] (1/4) Epoch 21, batch 7400, loss[loss=0.2592, simple_loss=0.3172, pruned_loss=0.1006, over 8290.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2885, pruned_loss=0.06299, over 1613200.22 frames. ], batch size: 23, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:59:53,401 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 03:00:10,729 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.483e+02 2.322e+02 3.020e+02 4.298e+02 1.187e+03, threshold=6.039e+02, percent-clipped=6.0 +2023-02-07 03:00:19,199 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169100.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:00:25,150 INFO [train.py:901] (1/4) Epoch 21, batch 7450, loss[loss=0.2248, simple_loss=0.3017, pruned_loss=0.07393, over 7019.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2877, pruned_loss=0.06227, over 1610901.25 frames. ], batch size: 72, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 03:00:33,888 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 03:00:42,572 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169134.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:01:01,173 INFO [train.py:901] (1/4) Epoch 21, batch 7500, loss[loss=0.2103, simple_loss=0.2906, pruned_loss=0.06495, over 7972.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2881, pruned_loss=0.06209, over 1614179.08 frames. ], batch size: 21, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 03:01:13,525 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:01:21,445 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.287e+02 2.739e+02 3.438e+02 5.948e+02, threshold=5.478e+02, percent-clipped=0.0 +2023-02-07 03:01:35,746 INFO [train.py:901] (1/4) Epoch 21, batch 7550, loss[loss=0.2, simple_loss=0.283, pruned_loss=0.05853, over 8256.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2872, pruned_loss=0.06169, over 1615842.66 frames. ], batch size: 24, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 03:02:09,759 INFO [train.py:901] (1/4) Epoch 21, batch 7600, loss[loss=0.202, simple_loss=0.2782, pruned_loss=0.06289, over 7439.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2871, pruned_loss=0.06138, over 1613359.44 frames. ], batch size: 17, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 03:02:32,179 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.243e+02 2.742e+02 3.349e+02 1.012e+03, threshold=5.485e+02, percent-clipped=5.0 +2023-02-07 03:02:45,876 INFO [train.py:901] (1/4) Epoch 21, batch 7650, loss[loss=0.1939, simple_loss=0.2878, pruned_loss=0.05, over 7974.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2872, pruned_loss=0.06141, over 1611659.75 frames. ], batch size: 21, lr: 3.57e-03, grad_scale: 16.0 +2023-02-07 03:03:00,453 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169329.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:03:01,835 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169331.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 03:03:21,298 INFO [train.py:901] (1/4) Epoch 21, batch 7700, loss[loss=0.206, simple_loss=0.2897, pruned_loss=0.06115, over 7810.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2876, pruned_loss=0.0616, over 1614457.40 frames. ], batch size: 20, lr: 3.57e-03, grad_scale: 16.0 +2023-02-07 03:03:37,725 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.51 vs. limit=5.0 +2023-02-07 03:03:42,194 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.349e+02 2.901e+02 3.736e+02 6.675e+02, threshold=5.802e+02, percent-clipped=6.0 +2023-02-07 03:03:44,253 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 03:03:45,054 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2556, 2.1063, 1.6375, 1.9647, 1.7497, 1.4104, 1.6720, 1.6901], + device='cuda:1'), covar=tensor([0.1265, 0.0380, 0.1142, 0.0524, 0.0727, 0.1455, 0.0885, 0.0872], + device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0236, 0.0335, 0.0307, 0.0301, 0.0337, 0.0345, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 03:03:57,021 INFO [train.py:901] (1/4) Epoch 21, batch 7750, loss[loss=0.1762, simple_loss=0.2661, pruned_loss=0.04319, over 8187.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2882, pruned_loss=0.06193, over 1614271.26 frames. ], batch size: 23, lr: 3.57e-03, grad_scale: 16.0 +2023-02-07 03:04:21,907 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169444.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:04:22,015 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169444.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:04:23,339 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169446.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 03:04:32,648 INFO [train.py:901] (1/4) Epoch 21, batch 7800, loss[loss=0.2205, simple_loss=0.2988, pruned_loss=0.07116, over 7714.00 frames. ], tot_loss[loss=0.206, simple_loss=0.288, pruned_loss=0.06199, over 1612553.48 frames. ], batch size: 18, lr: 3.57e-03, grad_scale: 16.0 +2023-02-07 03:04:45,392 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169478.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:04:52,675 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.145e+02 2.738e+02 3.428e+02 8.790e+02, threshold=5.476e+02, percent-clipped=3.0 +2023-02-07 03:05:06,025 INFO [train.py:901] (1/4) Epoch 21, batch 7850, loss[loss=0.1731, simple_loss=0.2581, pruned_loss=0.044, over 7925.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2875, pruned_loss=0.06149, over 1614576.40 frames. ], batch size: 20, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:05:14,138 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169521.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:05:39,266 INFO [train.py:901] (1/4) Epoch 21, batch 7900, loss[loss=0.1714, simple_loss=0.2478, pruned_loss=0.04746, over 7689.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2863, pruned_loss=0.06125, over 1610412.74 frames. ], batch size: 18, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:05:39,438 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169559.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:05:52,128 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2815, 1.2427, 3.4014, 1.0769, 3.0288, 2.8355, 3.1146, 3.0089], + device='cuda:1'), covar=tensor([0.0754, 0.4037, 0.0815, 0.4123, 0.1426, 0.1115, 0.0766, 0.0914], + device='cuda:1'), in_proj_covar=tensor([0.0631, 0.0641, 0.0695, 0.0627, 0.0711, 0.0612, 0.0612, 0.0676], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:05:59,284 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 2.405e+02 2.884e+02 3.520e+02 8.387e+02, threshold=5.767e+02, percent-clipped=5.0 +2023-02-07 03:06:02,045 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169593.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:06:11,880 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6411, 1.9302, 2.9421, 1.4991, 2.1334, 2.0388, 1.7018, 2.1887], + device='cuda:1'), covar=tensor([0.1810, 0.2573, 0.0895, 0.4573, 0.1901, 0.3112, 0.2357, 0.2222], + device='cuda:1'), in_proj_covar=tensor([0.0523, 0.0601, 0.0556, 0.0637, 0.0642, 0.0591, 0.0531, 0.0632], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:06:12,852 INFO [train.py:901] (1/4) Epoch 21, batch 7950, loss[loss=0.2009, simple_loss=0.2856, pruned_loss=0.05807, over 8462.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.288, pruned_loss=0.06212, over 1607076.70 frames. ], batch size: 25, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:06:31,343 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169636.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:06:33,349 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169639.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:06:46,591 INFO [train.py:901] (1/4) Epoch 21, batch 8000, loss[loss=0.203, simple_loss=0.2658, pruned_loss=0.07012, over 7691.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2879, pruned_loss=0.06216, over 1607677.58 frames. ], batch size: 18, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:07:06,442 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.455e+02 2.194e+02 2.844e+02 3.383e+02 6.688e+02, threshold=5.687e+02, percent-clipped=2.0 +2023-02-07 03:07:07,194 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2855, 3.2055, 2.9642, 1.4243, 2.9006, 2.9324, 2.9633, 2.8523], + device='cuda:1'), covar=tensor([0.1146, 0.0757, 0.1335, 0.4682, 0.1060, 0.1148, 0.1542, 0.1044], + device='cuda:1'), in_proj_covar=tensor([0.0521, 0.0428, 0.0431, 0.0531, 0.0422, 0.0440, 0.0421, 0.0380], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:07:12,045 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7861, 2.5630, 3.4169, 2.6742, 3.3352, 2.6742, 2.5334, 2.0563], + device='cuda:1'), covar=tensor([0.5401, 0.5397, 0.1912, 0.3862, 0.2570, 0.2922, 0.1817, 0.5641], + device='cuda:1'), in_proj_covar=tensor([0.0945, 0.0977, 0.0803, 0.0946, 0.0998, 0.0892, 0.0746, 0.0823], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 03:07:12,148 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 +2023-02-07 03:07:14,033 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169700.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:07:15,418 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169702.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 03:07:19,702 INFO [train.py:901] (1/4) Epoch 21, batch 8050, loss[loss=0.1992, simple_loss=0.2783, pruned_loss=0.06004, over 7244.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.286, pruned_loss=0.06206, over 1583637.07 frames. ], batch size: 16, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:07:30,628 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169725.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:07:32,003 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169727.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 03:07:53,204 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 03:07:58,218 INFO [train.py:901] (1/4) Epoch 22, batch 0, loss[loss=0.2471, simple_loss=0.3255, pruned_loss=0.08432, over 8358.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3255, pruned_loss=0.08432, over 8358.00 frames. ], batch size: 24, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:07:58,218 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 03:08:05,225 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6391, 1.7508, 1.5866, 1.9207, 1.3092, 1.5188, 1.7035, 1.7670], + device='cuda:1'), covar=tensor([0.0729, 0.0798, 0.0806, 0.0589, 0.1029, 0.1095, 0.0633, 0.0716], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0196, 0.0245, 0.0213, 0.0207, 0.0246, 0.0250, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 03:08:09,347 INFO [train.py:935] (1/4) Epoch 22, validation: loss=0.1743, simple_loss=0.2746, pruned_loss=0.03702, over 944034.00 frames. +2023-02-07 03:08:09,349 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 03:08:12,906 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:08:17,064 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9997, 1.4695, 1.6766, 1.4677, 0.8828, 1.4862, 1.7655, 1.6655], + device='cuda:1'), covar=tensor([0.0525, 0.1277, 0.1725, 0.1446, 0.0623, 0.1490, 0.0681, 0.0621], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0160, 0.0100, 0.0164, 0.0113, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 03:08:24,250 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 03:08:25,070 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169765.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:08:36,868 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5058, 2.4935, 1.9488, 2.2912, 2.2042, 1.6569, 2.0627, 2.1536], + device='cuda:1'), covar=tensor([0.1514, 0.0459, 0.1220, 0.0664, 0.0685, 0.1534, 0.0919, 0.0918], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0234, 0.0331, 0.0305, 0.0298, 0.0332, 0.0340, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 03:08:42,189 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.482e+02 2.980e+02 3.558e+02 1.069e+03, threshold=5.959e+02, percent-clipped=8.0 +2023-02-07 03:08:44,174 INFO [train.py:901] (1/4) Epoch 22, batch 50, loss[loss=0.1933, simple_loss=0.284, pruned_loss=0.05132, over 7964.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2928, pruned_loss=0.06392, over 364175.28 frames. ], batch size: 21, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:08:54,112 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169804.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:01,045 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 03:09:02,041 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:06,926 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2777, 2.0766, 2.6980, 2.2432, 2.6467, 2.3636, 2.1380, 1.4724], + device='cuda:1'), covar=tensor([0.5047, 0.4808, 0.1934, 0.3850, 0.2534, 0.3087, 0.1884, 0.5188], + device='cuda:1'), in_proj_covar=tensor([0.0943, 0.0975, 0.0801, 0.0943, 0.0997, 0.0889, 0.0744, 0.0822], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 03:09:19,148 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169840.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:20,323 INFO [train.py:901] (1/4) Epoch 22, batch 100, loss[loss=0.1985, simple_loss=0.2685, pruned_loss=0.06429, over 7938.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2917, pruned_loss=0.06333, over 643922.18 frames. ], batch size: 20, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:09:23,124 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 03:09:25,376 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169849.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:42,084 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169874.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:52,902 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.649e+02 2.356e+02 3.069e+02 3.800e+02 7.981e+02, threshold=6.138e+02, percent-clipped=3.0 +2023-02-07 03:09:55,641 INFO [train.py:901] (1/4) Epoch 22, batch 150, loss[loss=0.2254, simple_loss=0.3148, pruned_loss=0.06802, over 8567.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2916, pruned_loss=0.06311, over 862414.69 frames. ], batch size: 31, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:09:55,879 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169892.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:10:12,773 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169917.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:10:27,472 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5508, 1.4605, 2.8541, 1.3466, 2.1392, 3.0396, 3.1844, 2.5933], + device='cuda:1'), covar=tensor([0.1313, 0.1655, 0.0400, 0.2235, 0.1010, 0.0322, 0.0692, 0.0597], + device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0321, 0.0285, 0.0317, 0.0307, 0.0264, 0.0418, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 03:10:30,762 INFO [train.py:901] (1/4) Epoch 22, batch 200, loss[loss=0.195, simple_loss=0.2814, pruned_loss=0.05435, over 8354.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2903, pruned_loss=0.06216, over 1030705.19 frames. ], batch size: 24, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:10:53,647 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.98 vs. limit=5.0 +2023-02-07 03:10:58,691 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169983.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:11:02,622 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.362e+02 2.871e+02 3.395e+02 8.094e+02, threshold=5.742e+02, percent-clipped=2.0 +2023-02-07 03:11:04,633 INFO [train.py:901] (1/4) Epoch 22, batch 250, loss[loss=0.2414, simple_loss=0.3164, pruned_loss=0.08318, over 8435.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2897, pruned_loss=0.06193, over 1159810.23 frames. ], batch size: 27, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:11:17,872 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 03:11:26,104 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 03:11:41,667 INFO [train.py:901] (1/4) Epoch 22, batch 300, loss[loss=0.2676, simple_loss=0.3301, pruned_loss=0.1026, over 8100.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2877, pruned_loss=0.06112, over 1262067.71 frames. ], batch size: 23, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:11:56,574 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170063.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:12:13,700 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.486e+02 2.821e+02 3.492e+02 6.452e+02, threshold=5.641e+02, percent-clipped=3.0 +2023-02-07 03:12:15,181 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170091.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:12:15,775 INFO [train.py:901] (1/4) Epoch 22, batch 350, loss[loss=0.2236, simple_loss=0.2993, pruned_loss=0.07397, over 8134.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2886, pruned_loss=0.062, over 1335642.63 frames. ], batch size: 22, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:12:19,956 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170098.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:12:27,049 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170109.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:12:49,716 INFO [train.py:901] (1/4) Epoch 22, batch 400, loss[loss=0.2135, simple_loss=0.3015, pruned_loss=0.06274, over 8606.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2878, pruned_loss=0.06143, over 1396804.01 frames. ], batch size: 34, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:12:53,693 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170148.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:13:22,572 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.277e+02 2.821e+02 3.460e+02 6.418e+02, threshold=5.643e+02, percent-clipped=3.0 +2023-02-07 03:13:24,661 INFO [train.py:901] (1/4) Epoch 22, batch 450, loss[loss=0.1919, simple_loss=0.2662, pruned_loss=0.05886, over 7689.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2881, pruned_loss=0.06158, over 1446951.56 frames. ], batch size: 18, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:13:34,381 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170206.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:13:46,334 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170224.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:13:58,272 INFO [train.py:901] (1/4) Epoch 22, batch 500, loss[loss=0.1693, simple_loss=0.2428, pruned_loss=0.04787, over 7420.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2885, pruned_loss=0.06162, over 1487645.88 frames. ], batch size: 17, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:14:13,733 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170263.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:14:31,688 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.263e+02 2.770e+02 3.716e+02 6.957e+02, threshold=5.540e+02, percent-clipped=5.0 +2023-02-07 03:14:34,527 INFO [train.py:901] (1/4) Epoch 22, batch 550, loss[loss=0.186, simple_loss=0.28, pruned_loss=0.04601, over 8478.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2887, pruned_loss=0.06248, over 1514854.28 frames. ], batch size: 28, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:15:00,823 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0143, 2.2607, 1.8518, 2.6694, 1.4924, 1.6714, 2.1082, 2.2067], + device='cuda:1'), covar=tensor([0.0741, 0.0783, 0.0956, 0.0487, 0.1156, 0.1317, 0.0799, 0.0797], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0196, 0.0243, 0.0213, 0.0206, 0.0245, 0.0248, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 03:15:08,214 INFO [train.py:901] (1/4) Epoch 22, batch 600, loss[loss=0.2017, simple_loss=0.2878, pruned_loss=0.05781, over 8545.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.289, pruned_loss=0.06219, over 1541352.01 frames. ], batch size: 31, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:15:16,556 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170354.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:15:27,507 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 03:15:34,303 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170379.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:15:40,800 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.463e+02 3.010e+02 3.561e+02 9.437e+02, threshold=6.021e+02, percent-clipped=1.0 +2023-02-07 03:15:42,758 INFO [train.py:901] (1/4) Epoch 22, batch 650, loss[loss=0.1962, simple_loss=0.2793, pruned_loss=0.05651, over 8575.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2899, pruned_loss=0.0628, over 1558988.36 frames. ], batch size: 34, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:15:52,768 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.2468, 5.2527, 4.7383, 2.5481, 4.6054, 4.9639, 4.8895, 4.6986], + device='cuda:1'), covar=tensor([0.0562, 0.0393, 0.0790, 0.4262, 0.0782, 0.0926, 0.1109, 0.0679], + device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0430, 0.0428, 0.0530, 0.0422, 0.0441, 0.0423, 0.0381], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:15:53,432 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170407.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:16:17,635 INFO [train.py:901] (1/4) Epoch 22, batch 700, loss[loss=0.1809, simple_loss=0.2609, pruned_loss=0.05045, over 7702.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2877, pruned_loss=0.06169, over 1570588.86 frames. ], batch size: 18, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:16:31,480 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170462.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:16:42,963 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170479.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:16:43,712 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170480.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:16:49,127 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170487.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:16:50,917 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.347e+02 2.936e+02 3.672e+02 5.936e+02, threshold=5.871e+02, percent-clipped=0.0 +2023-02-07 03:16:52,903 INFO [train.py:901] (1/4) Epoch 22, batch 750, loss[loss=0.2038, simple_loss=0.2903, pruned_loss=0.05861, over 8353.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2872, pruned_loss=0.06173, over 1580181.33 frames. ], batch size: 24, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:17:01,715 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170505.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:11,628 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170519.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:13,545 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170522.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:14,728 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 03:17:23,963 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 03:17:27,405 INFO [train.py:901] (1/4) Epoch 22, batch 800, loss[loss=0.2007, simple_loss=0.281, pruned_loss=0.06015, over 7967.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2874, pruned_loss=0.06186, over 1590613.63 frames. ], batch size: 21, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:17:28,966 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170544.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:30,549 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 +2023-02-07 03:17:57,572 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170587.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:58,755 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 2.234e+02 2.598e+02 3.180e+02 6.753e+02, threshold=5.195e+02, percent-clipped=1.0 +2023-02-07 03:18:00,808 INFO [train.py:901] (1/4) Epoch 22, batch 850, loss[loss=0.1651, simple_loss=0.2584, pruned_loss=0.03594, over 8083.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2869, pruned_loss=0.06166, over 1597193.41 frames. ], batch size: 21, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:18:14,723 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8591, 2.4092, 3.6746, 1.9219, 1.8696, 3.7249, 0.7287, 2.1679], + device='cuda:1'), covar=tensor([0.1577, 0.1371, 0.0257, 0.1890, 0.2993, 0.0322, 0.2508, 0.1465], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0196, 0.0127, 0.0222, 0.0270, 0.0135, 0.0171, 0.0193], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 03:18:31,335 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-07 03:18:36,994 INFO [train.py:901] (1/4) Epoch 22, batch 900, loss[loss=0.2164, simple_loss=0.3028, pruned_loss=0.06499, over 8506.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2857, pruned_loss=0.06092, over 1600196.55 frames. ], batch size: 28, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:18:53,614 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2512, 2.5153, 2.9899, 1.7001, 3.1308, 1.9449, 1.6264, 2.1069], + device='cuda:1'), covar=tensor([0.0720, 0.0432, 0.0269, 0.0729, 0.0406, 0.0843, 0.0883, 0.0529], + device='cuda:1'), in_proj_covar=tensor([0.0454, 0.0394, 0.0345, 0.0444, 0.0375, 0.0534, 0.0390, 0.0419], + device='cuda:1'), out_proj_covar=tensor([1.2187e-04, 1.0338e-04, 9.0653e-05, 1.1683e-04, 9.8477e-05, 1.5078e-04, + 1.0542e-04, 1.1114e-04], device='cuda:1') +2023-02-07 03:19:09,381 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.381e+02 2.827e+02 3.296e+02 7.509e+02, threshold=5.655e+02, percent-clipped=4.0 +2023-02-07 03:19:11,448 INFO [train.py:901] (1/4) Epoch 22, batch 950, loss[loss=0.2178, simple_loss=0.3017, pruned_loss=0.06693, over 8482.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2856, pruned_loss=0.0608, over 1603122.86 frames. ], batch size: 29, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:19:13,662 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9379, 1.5737, 3.1366, 1.6085, 2.2849, 3.4094, 3.5001, 2.9492], + device='cuda:1'), covar=tensor([0.1087, 0.1635, 0.0323, 0.1892, 0.0897, 0.0236, 0.0511, 0.0514], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0323, 0.0286, 0.0317, 0.0310, 0.0265, 0.0420, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 03:19:43,603 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 03:19:46,359 INFO [train.py:901] (1/4) Epoch 22, batch 1000, loss[loss=0.2087, simple_loss=0.2821, pruned_loss=0.06767, over 8488.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2861, pruned_loss=0.06132, over 1605641.34 frames. ], batch size: 25, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:19:49,226 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2380, 2.1197, 1.6969, 1.9356, 1.7783, 1.4641, 1.6766, 1.6584], + device='cuda:1'), covar=tensor([0.1257, 0.0421, 0.1180, 0.0501, 0.0719, 0.1381, 0.0933, 0.0922], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0235, 0.0331, 0.0308, 0.0299, 0.0334, 0.0341, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 03:20:12,115 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170778.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:20:17,087 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 03:20:19,772 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.250e+02 2.326e+02 2.890e+02 3.504e+02 6.405e+02, threshold=5.779e+02, percent-clipped=4.0 +2023-02-07 03:20:21,021 INFO [train.py:901] (1/4) Epoch 22, batch 1050, loss[loss=0.2237, simple_loss=0.3026, pruned_loss=0.07237, over 8506.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.286, pruned_loss=0.0614, over 1606797.38 frames. ], batch size: 26, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:20:28,506 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 03:20:28,713 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170803.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:20:41,785 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170823.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:20:55,958 INFO [train.py:901] (1/4) Epoch 22, batch 1100, loss[loss=0.2447, simple_loss=0.3343, pruned_loss=0.07752, over 8453.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.287, pruned_loss=0.06186, over 1609538.13 frames. ], batch size: 27, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:21:27,508 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2684, 2.1135, 1.6839, 1.8945, 1.8070, 1.4494, 1.6995, 1.6018], + device='cuda:1'), covar=tensor([0.1375, 0.0421, 0.1159, 0.0555, 0.0708, 0.1404, 0.0930, 0.0936], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0237, 0.0334, 0.0310, 0.0300, 0.0336, 0.0344, 0.0319], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 03:21:29,306 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.559e+02 2.576e+02 3.127e+02 3.706e+02 1.049e+03, threshold=6.255e+02, percent-clipped=5.0 +2023-02-07 03:21:30,677 INFO [train.py:901] (1/4) Epoch 22, batch 1150, loss[loss=0.2213, simple_loss=0.3035, pruned_loss=0.06957, over 8506.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2884, pruned_loss=0.06242, over 1617329.61 frames. ], batch size: 26, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:21:37,421 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 03:21:45,384 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5561, 1.5132, 2.6349, 1.1268, 2.0772, 2.9031, 3.1532, 2.1037], + device='cuda:1'), covar=tensor([0.1558, 0.1842, 0.0578, 0.2826, 0.1158, 0.0417, 0.0719, 0.1086], + device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0320, 0.0283, 0.0314, 0.0306, 0.0262, 0.0415, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 03:21:52,865 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0195, 2.1731, 1.8455, 2.7279, 1.3495, 1.6268, 1.9697, 2.2624], + device='cuda:1'), covar=tensor([0.0720, 0.0821, 0.0896, 0.0383, 0.1075, 0.1315, 0.0794, 0.0659], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0198, 0.0247, 0.0215, 0.0209, 0.0249, 0.0251, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 03:21:56,780 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170931.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:22:01,710 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170938.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:22:04,213 INFO [train.py:901] (1/4) Epoch 22, batch 1200, loss[loss=0.225, simple_loss=0.3011, pruned_loss=0.07445, over 8186.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2881, pruned_loss=0.0623, over 1619736.58 frames. ], batch size: 23, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:22:07,064 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170946.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:22:38,801 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.384e+02 2.807e+02 3.549e+02 5.873e+02, threshold=5.615e+02, percent-clipped=0.0 +2023-02-07 03:22:40,092 INFO [train.py:901] (1/4) Epoch 22, batch 1250, loss[loss=0.1839, simple_loss=0.2766, pruned_loss=0.04558, over 8352.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2871, pruned_loss=0.06143, over 1615037.45 frames. ], batch size: 24, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:22:57,667 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5125, 4.4737, 4.0141, 2.1417, 3.8498, 4.1398, 4.1789, 3.8919], + device='cuda:1'), covar=tensor([0.0771, 0.0589, 0.1059, 0.4750, 0.0976, 0.1077, 0.1152, 0.0796], + device='cuda:1'), in_proj_covar=tensor([0.0522, 0.0430, 0.0430, 0.0530, 0.0422, 0.0441, 0.0420, 0.0382], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:22:58,786 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 03:23:14,555 INFO [train.py:901] (1/4) Epoch 22, batch 1300, loss[loss=0.2578, simple_loss=0.3252, pruned_loss=0.09514, over 6760.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2868, pruned_loss=0.06085, over 1616737.61 frames. ], batch size: 71, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:23:17,502 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171046.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:23:47,495 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.382e+02 2.988e+02 3.753e+02 7.309e+02, threshold=5.975e+02, percent-clipped=5.0 +2023-02-07 03:23:48,841 INFO [train.py:901] (1/4) Epoch 22, batch 1350, loss[loss=0.2133, simple_loss=0.3024, pruned_loss=0.06208, over 8254.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.288, pruned_loss=0.06167, over 1620504.28 frames. ], batch size: 24, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:24:01,667 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171110.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:24:02,353 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1748, 1.3040, 3.3204, 1.0699, 2.9269, 2.7552, 3.0272, 2.9057], + device='cuda:1'), covar=tensor([0.0890, 0.4332, 0.0918, 0.4340, 0.1510, 0.1256, 0.0828, 0.1000], + device='cuda:1'), in_proj_covar=tensor([0.0625, 0.0637, 0.0688, 0.0620, 0.0704, 0.0604, 0.0606, 0.0672], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:24:23,452 INFO [train.py:901] (1/4) Epoch 22, batch 1400, loss[loss=0.2191, simple_loss=0.3006, pruned_loss=0.06883, over 8294.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2887, pruned_loss=0.06221, over 1621356.48 frames. ], batch size: 23, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:24:23,825 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.05 vs. limit=5.0 +2023-02-07 03:24:55,486 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.434e+02 3.047e+02 3.835e+02 9.203e+02, threshold=6.094e+02, percent-clipped=3.0 +2023-02-07 03:24:57,482 INFO [train.py:901] (1/4) Epoch 22, batch 1450, loss[loss=0.217, simple_loss=0.291, pruned_loss=0.07145, over 8332.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2896, pruned_loss=0.06294, over 1618460.21 frames. ], batch size: 25, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:24:58,895 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171194.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:25:06,229 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 03:25:12,520 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:25:16,689 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171219.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:25:32,454 INFO [train.py:901] (1/4) Epoch 22, batch 1500, loss[loss=0.225, simple_loss=0.2929, pruned_loss=0.07855, over 7811.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.29, pruned_loss=0.06358, over 1618441.69 frames. ], batch size: 20, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:26:04,592 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.477e+02 2.962e+02 3.885e+02 1.079e+03, threshold=5.924e+02, percent-clipped=2.0 +2023-02-07 03:26:04,685 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=171290.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:26:05,961 INFO [train.py:901] (1/4) Epoch 22, batch 1550, loss[loss=0.1902, simple_loss=0.2817, pruned_loss=0.04937, over 7251.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2908, pruned_loss=0.06392, over 1619739.90 frames. ], batch size: 16, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:26:12,941 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171302.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:26:25,491 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7606, 1.9356, 1.6332, 2.2656, 1.0117, 1.4695, 1.6354, 1.9641], + device='cuda:1'), covar=tensor([0.0730, 0.0727, 0.0936, 0.0442, 0.1157, 0.1313, 0.0823, 0.0669], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0196, 0.0244, 0.0214, 0.0207, 0.0247, 0.0250, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 03:26:30,108 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171327.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:26:37,455 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([6.0244, 1.5092, 6.2001, 2.3016, 5.6462, 5.2362, 5.7144, 5.5971], + device='cuda:1'), covar=tensor([0.0346, 0.4680, 0.0320, 0.3655, 0.0812, 0.0821, 0.0411, 0.0419], + device='cuda:1'), in_proj_covar=tensor([0.0629, 0.0642, 0.0693, 0.0623, 0.0708, 0.0608, 0.0611, 0.0676], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:26:40,682 INFO [train.py:901] (1/4) Epoch 22, batch 1600, loss[loss=0.1923, simple_loss=0.2785, pruned_loss=0.05311, over 8285.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2905, pruned_loss=0.06342, over 1623080.31 frames. ], batch size: 23, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:26:55,765 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171363.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:27:13,639 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.350e+02 2.510e+02 3.045e+02 3.987e+02 6.104e+02, threshold=6.090e+02, percent-clipped=2.0 +2023-02-07 03:27:15,008 INFO [train.py:901] (1/4) Epoch 22, batch 1650, loss[loss=0.1957, simple_loss=0.2847, pruned_loss=0.05331, over 8102.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2914, pruned_loss=0.06353, over 1625997.78 frames. ], batch size: 23, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:27:24,118 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171405.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:27:51,035 INFO [train.py:901] (1/4) Epoch 22, batch 1700, loss[loss=0.2, simple_loss=0.2803, pruned_loss=0.05981, over 8665.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2906, pruned_loss=0.06313, over 1621930.91 frames. ], batch size: 34, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:27:59,275 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=171454.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:28:18,869 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-07 03:28:24,558 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.429e+02 3.050e+02 3.629e+02 7.357e+02, threshold=6.100e+02, percent-clipped=3.0 +2023-02-07 03:28:25,928 INFO [train.py:901] (1/4) Epoch 22, batch 1750, loss[loss=0.2195, simple_loss=0.3074, pruned_loss=0.06575, over 8465.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2906, pruned_loss=0.06287, over 1622478.53 frames. ], batch size: 27, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:28:42,131 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171516.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:29:00,471 INFO [train.py:901] (1/4) Epoch 22, batch 1800, loss[loss=0.2014, simple_loss=0.2879, pruned_loss=0.05751, over 8704.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2901, pruned_loss=0.06261, over 1625070.42 frames. ], batch size: 30, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:29:11,512 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=171558.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:29:19,688 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171569.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:29:34,591 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.408e+02 2.801e+02 3.784e+02 7.831e+02, threshold=5.602e+02, percent-clipped=2.0 +2023-02-07 03:29:35,959 INFO [train.py:901] (1/4) Epoch 22, batch 1850, loss[loss=0.185, simple_loss=0.2615, pruned_loss=0.05428, over 8086.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2897, pruned_loss=0.06254, over 1621797.36 frames. ], batch size: 21, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:30:10,050 INFO [train.py:901] (1/4) Epoch 22, batch 1900, loss[loss=0.1668, simple_loss=0.2489, pruned_loss=0.04233, over 7797.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2896, pruned_loss=0.06233, over 1615742.66 frames. ], batch size: 19, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:30:24,230 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171661.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:30:32,195 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171673.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:30:36,782 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 03:30:41,590 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171686.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:30:44,068 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.518e+02 3.035e+02 3.649e+02 9.576e+02, threshold=6.070e+02, percent-clipped=4.0 +2023-02-07 03:30:45,460 INFO [train.py:901] (1/4) Epoch 22, batch 1950, loss[loss=0.185, simple_loss=0.2631, pruned_loss=0.05343, over 7974.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2897, pruned_loss=0.06243, over 1620229.78 frames. ], batch size: 21, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:30:48,014 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 03:30:56,295 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=171707.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:31:07,796 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 03:31:20,038 INFO [train.py:901] (1/4) Epoch 22, batch 2000, loss[loss=0.187, simple_loss=0.2719, pruned_loss=0.05104, over 7932.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2885, pruned_loss=0.06202, over 1614514.30 frames. ], batch size: 20, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:31:43,119 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0720, 1.5745, 3.4401, 1.6514, 2.3616, 3.7944, 3.8678, 3.2127], + device='cuda:1'), covar=tensor([0.1155, 0.1822, 0.0353, 0.2141, 0.1153, 0.0207, 0.0589, 0.0574], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0323, 0.0285, 0.0318, 0.0307, 0.0264, 0.0419, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 03:31:53,990 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.301e+02 2.928e+02 3.706e+02 6.798e+02, threshold=5.855e+02, percent-clipped=1.0 +2023-02-07 03:31:55,401 INFO [train.py:901] (1/4) Epoch 22, batch 2050, loss[loss=0.2532, simple_loss=0.3318, pruned_loss=0.08729, over 8248.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2883, pruned_loss=0.06227, over 1607288.90 frames. ], batch size: 24, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:32:17,606 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171822.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:32:19,738 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171825.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:32:31,146 INFO [train.py:901] (1/4) Epoch 22, batch 2100, loss[loss=0.1748, simple_loss=0.2571, pruned_loss=0.04625, over 7778.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2881, pruned_loss=0.06197, over 1609001.04 frames. ], batch size: 19, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:32:36,878 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171850.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:32:43,494 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=171860.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:32:45,802 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5245, 1.5454, 2.0860, 1.4382, 1.1734, 2.0612, 0.3649, 1.2546], + device='cuda:1'), covar=tensor([0.1838, 0.1187, 0.0413, 0.1125, 0.2646, 0.0416, 0.2153, 0.1293], + device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0198, 0.0127, 0.0223, 0.0272, 0.0137, 0.0171, 0.0194], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 03:33:05,589 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.505e+02 2.999e+02 3.749e+02 9.868e+02, threshold=5.998e+02, percent-clipped=7.0 +2023-02-07 03:33:06,896 INFO [train.py:901] (1/4) Epoch 22, batch 2150, loss[loss=0.2133, simple_loss=0.2924, pruned_loss=0.06711, over 7813.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2878, pruned_loss=0.0615, over 1609961.38 frames. ], batch size: 20, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:33:33,122 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171929.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:33:42,595 INFO [train.py:901] (1/4) Epoch 22, batch 2200, loss[loss=0.2273, simple_loss=0.3091, pruned_loss=0.0728, over 8226.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2881, pruned_loss=0.06191, over 1609731.35 frames. ], batch size: 22, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:33:51,045 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171954.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:33:51,683 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171955.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:34:04,374 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 +2023-02-07 03:34:05,587 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171975.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:34:15,550 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.362e+02 2.812e+02 3.623e+02 6.076e+02, threshold=5.624e+02, percent-clipped=1.0 +2023-02-07 03:34:16,938 INFO [train.py:901] (1/4) Epoch 22, batch 2250, loss[loss=0.1706, simple_loss=0.2521, pruned_loss=0.04459, over 7815.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2872, pruned_loss=0.06135, over 1609549.05 frames. ], batch size: 20, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:34:17,296 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-07 03:34:54,268 INFO [train.py:901] (1/4) Epoch 22, batch 2300, loss[loss=0.2363, simple_loss=0.3206, pruned_loss=0.07601, over 8339.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2878, pruned_loss=0.06164, over 1605704.81 frames. ], batch size: 26, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:35:19,362 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6009, 2.4709, 1.8756, 2.3157, 2.1505, 1.6317, 2.0654, 2.1384], + device='cuda:1'), covar=tensor([0.1310, 0.0373, 0.1088, 0.0517, 0.0723, 0.1421, 0.0904, 0.0824], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0233, 0.0331, 0.0307, 0.0299, 0.0336, 0.0343, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 03:35:20,118 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172078.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:35:28,298 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.401e+02 3.005e+02 3.667e+02 7.010e+02, threshold=6.010e+02, percent-clipped=1.0 +2023-02-07 03:35:29,621 INFO [train.py:901] (1/4) Epoch 22, batch 2350, loss[loss=0.1963, simple_loss=0.2648, pruned_loss=0.0639, over 7714.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2869, pruned_loss=0.06125, over 1606186.35 frames. ], batch size: 18, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:35:37,284 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172103.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:36:01,302 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172136.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:36:05,211 INFO [train.py:901] (1/4) Epoch 22, batch 2400, loss[loss=0.1892, simple_loss=0.2838, pruned_loss=0.04724, over 8456.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2875, pruned_loss=0.06187, over 1605935.83 frames. ], batch size: 25, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:36:39,685 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.658e+02 3.455e+02 4.348e+02 7.809e+02, threshold=6.910e+02, percent-clipped=6.0 +2023-02-07 03:36:41,123 INFO [train.py:901] (1/4) Epoch 22, batch 2450, loss[loss=0.2304, simple_loss=0.3056, pruned_loss=0.07763, over 8526.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.288, pruned_loss=0.06235, over 1609310.56 frames. ], batch size: 49, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:36:45,750 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1862, 1.8419, 2.3585, 2.0868, 2.3908, 2.1470, 1.9364, 1.1900], + device='cuda:1'), covar=tensor([0.4837, 0.4665, 0.1865, 0.3323, 0.1977, 0.3143, 0.1841, 0.4775], + device='cuda:1'), in_proj_covar=tensor([0.0944, 0.0978, 0.0805, 0.0942, 0.0997, 0.0895, 0.0748, 0.0828], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 03:37:08,131 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172231.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:37:16,956 INFO [train.py:901] (1/4) Epoch 22, batch 2500, loss[loss=0.2067, simple_loss=0.3071, pruned_loss=0.05311, over 8526.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2874, pruned_loss=0.06143, over 1615174.70 frames. ], batch size: 49, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:37:26,696 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172256.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:37:37,003 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8294, 1.9671, 1.7670, 2.6155, 1.1899, 1.5663, 1.7906, 1.9784], + device='cuda:1'), covar=tensor([0.0764, 0.0766, 0.0889, 0.0367, 0.1049, 0.1270, 0.0851, 0.0809], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0197, 0.0243, 0.0215, 0.0206, 0.0248, 0.0251, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 03:37:50,845 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.288e+02 2.722e+02 3.540e+02 9.975e+02, threshold=5.443e+02, percent-clipped=1.0 +2023-02-07 03:37:52,252 INFO [train.py:901] (1/4) Epoch 22, batch 2550, loss[loss=0.1954, simple_loss=0.2749, pruned_loss=0.05791, over 8141.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2873, pruned_loss=0.06175, over 1614594.94 frames. ], batch size: 22, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:37:56,720 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=172299.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:38:25,419 INFO [train.py:901] (1/4) Epoch 22, batch 2600, loss[loss=0.1982, simple_loss=0.2873, pruned_loss=0.0546, over 8383.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2862, pruned_loss=0.06128, over 1612761.78 frames. ], batch size: 48, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:38:58,398 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.484e+02 3.096e+02 3.957e+02 1.134e+03, threshold=6.191e+02, percent-clipped=6.0 +2023-02-07 03:39:00,470 INFO [train.py:901] (1/4) Epoch 22, batch 2650, loss[loss=0.1832, simple_loss=0.2683, pruned_loss=0.04901, over 8460.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.286, pruned_loss=0.06134, over 1617202.20 frames. ], batch size: 25, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:39:16,288 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172414.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:39:35,340 INFO [train.py:901] (1/4) Epoch 22, batch 2700, loss[loss=0.201, simple_loss=0.2904, pruned_loss=0.05577, over 8331.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2861, pruned_loss=0.06146, over 1616731.23 frames. ], batch size: 25, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:40:02,834 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=172480.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:40:09,203 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-07 03:40:09,436 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 2.345e+02 2.798e+02 3.767e+02 1.133e+03, threshold=5.596e+02, percent-clipped=4.0 +2023-02-07 03:40:10,848 INFO [train.py:901] (1/4) Epoch 22, batch 2750, loss[loss=0.2027, simple_loss=0.2896, pruned_loss=0.05791, over 8133.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2861, pruned_loss=0.06182, over 1615069.72 frames. ], batch size: 22, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:40:45,669 INFO [train.py:901] (1/4) Epoch 22, batch 2800, loss[loss=0.2106, simple_loss=0.2812, pruned_loss=0.06995, over 7196.00 frames. ], tot_loss[loss=0.205, simple_loss=0.286, pruned_loss=0.06196, over 1609799.44 frames. ], batch size: 16, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:41:18,226 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.794e+02 2.395e+02 2.840e+02 3.614e+02 7.820e+02, threshold=5.680e+02, percent-clipped=6.0 +2023-02-07 03:41:20,379 INFO [train.py:901] (1/4) Epoch 22, batch 2850, loss[loss=0.2411, simple_loss=0.3185, pruned_loss=0.08186, over 8301.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2866, pruned_loss=0.06203, over 1606579.40 frames. ], batch size: 23, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:41:23,220 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172595.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:41:37,768 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172616.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:41:56,013 INFO [train.py:901] (1/4) Epoch 22, batch 2900, loss[loss=0.1986, simple_loss=0.2932, pruned_loss=0.05203, over 8251.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2854, pruned_loss=0.06138, over 1602941.54 frames. ], batch size: 24, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:41:57,545 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:42:15,805 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172670.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:42:24,172 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 03:42:28,908 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.482e+02 2.975e+02 3.907e+02 6.756e+02, threshold=5.949e+02, percent-clipped=4.0 +2023-02-07 03:42:30,289 INFO [train.py:901] (1/4) Epoch 22, batch 2950, loss[loss=0.194, simple_loss=0.2791, pruned_loss=0.05448, over 7967.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2857, pruned_loss=0.06156, over 1605788.49 frames. ], batch size: 21, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:42:32,556 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172695.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:43:05,665 INFO [train.py:901] (1/4) Epoch 22, batch 3000, loss[loss=0.1782, simple_loss=0.2575, pruned_loss=0.0495, over 7645.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2854, pruned_loss=0.06133, over 1604961.10 frames. ], batch size: 19, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:43:05,665 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 03:43:17,970 INFO [train.py:935] (1/4) Epoch 22, validation: loss=0.1735, simple_loss=0.2739, pruned_loss=0.03659, over 944034.00 frames. +2023-02-07 03:43:17,971 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 03:43:25,618 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172752.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:43:51,446 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.191e+02 2.765e+02 3.574e+02 6.067e+02, threshold=5.530e+02, percent-clipped=1.0 +2023-02-07 03:43:52,760 INFO [train.py:901] (1/4) Epoch 22, batch 3050, loss[loss=0.1999, simple_loss=0.2812, pruned_loss=0.05927, over 8125.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2858, pruned_loss=0.06111, over 1606813.36 frames. ], batch size: 22, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:44:15,791 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7328, 1.3408, 3.3524, 1.4794, 2.2796, 3.7386, 3.9055, 3.1613], + device='cuda:1'), covar=tensor([0.1260, 0.1969, 0.0363, 0.2099, 0.1140, 0.0231, 0.0500, 0.0602], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0325, 0.0286, 0.0318, 0.0310, 0.0266, 0.0423, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 03:44:26,397 INFO [train.py:901] (1/4) Epoch 22, batch 3100, loss[loss=0.1844, simple_loss=0.2741, pruned_loss=0.0473, over 8196.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2849, pruned_loss=0.06082, over 1605899.67 frames. ], batch size: 23, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:44:31,357 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0154, 2.4542, 2.6836, 1.5091, 3.0550, 1.6585, 1.5047, 2.0922], + device='cuda:1'), covar=tensor([0.0910, 0.0402, 0.0314, 0.0849, 0.0456, 0.1000, 0.0854, 0.0579], + device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0386, 0.0340, 0.0440, 0.0370, 0.0528, 0.0385, 0.0412], + device='cuda:1'), out_proj_covar=tensor([1.1964e-04, 1.0113e-04, 8.9532e-05, 1.1579e-04, 9.7151e-05, 1.4874e-04, + 1.0389e-04, 1.0921e-04], device='cuda:1') +2023-02-07 03:44:32,686 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172851.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:44:40,603 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172863.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:44:44,821 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0541, 2.3042, 1.7611, 2.8844, 1.3553, 1.6739, 1.9948, 2.2796], + device='cuda:1'), covar=tensor([0.0717, 0.0652, 0.0906, 0.0328, 0.1011, 0.1155, 0.0751, 0.0756], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0198, 0.0246, 0.0217, 0.0208, 0.0249, 0.0252, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 03:44:47,321 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2455, 1.3031, 3.3492, 1.0997, 2.9370, 2.7920, 3.0658, 2.9613], + device='cuda:1'), covar=tensor([0.0762, 0.4433, 0.0801, 0.4223, 0.1420, 0.1160, 0.0825, 0.0879], + device='cuda:1'), in_proj_covar=tensor([0.0627, 0.0640, 0.0689, 0.0621, 0.0706, 0.0611, 0.0607, 0.0673], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:44:50,783 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172876.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:44:59,808 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.451e+02 3.163e+02 4.463e+02 7.617e+02, threshold=6.327e+02, percent-clipped=7.0 +2023-02-07 03:45:01,202 INFO [train.py:901] (1/4) Epoch 22, batch 3150, loss[loss=0.2341, simple_loss=0.3072, pruned_loss=0.08054, over 8333.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2868, pruned_loss=0.06191, over 1611160.86 frames. ], batch size: 26, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:45:27,822 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.33 vs. limit=5.0 +2023-02-07 03:45:29,139 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.94 vs. limit=5.0 +2023-02-07 03:45:35,480 INFO [train.py:901] (1/4) Epoch 22, batch 3200, loss[loss=0.1475, simple_loss=0.2277, pruned_loss=0.03359, over 7442.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2867, pruned_loss=0.06159, over 1613664.97 frames. ], batch size: 17, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:45:47,724 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=172960.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:45:53,342 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6464, 3.1149, 2.5384, 4.1476, 1.8036, 2.1934, 2.4287, 3.0904], + device='cuda:1'), covar=tensor([0.0591, 0.0614, 0.0740, 0.0204, 0.1004, 0.1122, 0.0889, 0.0744], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0198, 0.0246, 0.0217, 0.0207, 0.0249, 0.0252, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 03:46:06,752 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=172987.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:46:07,595 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9588, 1.7117, 2.0801, 1.8878, 2.0365, 1.9868, 1.8265, 0.8343], + device='cuda:1'), covar=tensor([0.5278, 0.4345, 0.1866, 0.3046, 0.2092, 0.2743, 0.1768, 0.4623], + device='cuda:1'), in_proj_covar=tensor([0.0945, 0.0978, 0.0805, 0.0944, 0.0997, 0.0896, 0.0748, 0.0828], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 03:46:09,262 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.554e+02 2.964e+02 3.773e+02 6.891e+02, threshold=5.928e+02, percent-clipped=2.0 +2023-02-07 03:46:10,572 INFO [train.py:901] (1/4) Epoch 22, batch 3250, loss[loss=0.2011, simple_loss=0.2749, pruned_loss=0.06366, over 7552.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2878, pruned_loss=0.062, over 1615229.46 frames. ], batch size: 18, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:46:19,610 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4163, 1.6642, 1.6649, 1.1925, 1.7297, 1.3423, 0.2942, 1.6285], + device='cuda:1'), covar=tensor([0.0428, 0.0332, 0.0275, 0.0440, 0.0323, 0.0763, 0.0859, 0.0225], + device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0388, 0.0342, 0.0443, 0.0372, 0.0531, 0.0387, 0.0414], + device='cuda:1'), out_proj_covar=tensor([1.2008e-04, 1.0164e-04, 9.0147e-05, 1.1659e-04, 9.7624e-05, 1.4961e-04, + 1.0443e-04, 1.0972e-04], device='cuda:1') +2023-02-07 03:46:40,192 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6269, 1.4967, 1.7628, 1.4205, 0.9150, 1.5636, 1.5517, 1.3219], + device='cuda:1'), covar=tensor([0.0546, 0.1218, 0.1519, 0.1371, 0.0590, 0.1401, 0.0676, 0.0687], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0158, 0.0099, 0.0163, 0.0111, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 03:46:45,374 INFO [train.py:901] (1/4) Epoch 22, batch 3300, loss[loss=0.1862, simple_loss=0.2665, pruned_loss=0.05298, over 7978.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2878, pruned_loss=0.0622, over 1613696.92 frames. ], batch size: 21, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:46:45,483 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8873, 6.1160, 5.3613, 2.6606, 5.4353, 5.7060, 5.5739, 5.4446], + device='cuda:1'), covar=tensor([0.0555, 0.0366, 0.0866, 0.4213, 0.0790, 0.0845, 0.1086, 0.0629], + device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0434, 0.0431, 0.0538, 0.0426, 0.0446, 0.0426, 0.0385], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:47:07,560 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173075.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:47:17,931 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.506e+02 2.831e+02 3.669e+02 6.075e+02, threshold=5.662e+02, percent-clipped=1.0 +2023-02-07 03:47:18,592 INFO [train.py:901] (1/4) Epoch 22, batch 3350, loss[loss=0.2618, simple_loss=0.3331, pruned_loss=0.09528, over 8285.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2884, pruned_loss=0.0627, over 1612045.04 frames. ], batch size: 23, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:47:22,022 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173096.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:47:23,438 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7018, 5.8880, 5.1273, 2.6023, 5.0917, 5.5184, 5.4152, 5.2461], + device='cuda:1'), covar=tensor([0.0620, 0.0444, 0.0959, 0.4465, 0.0828, 0.0896, 0.1113, 0.0630], + device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0434, 0.0432, 0.0536, 0.0426, 0.0446, 0.0425, 0.0385], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:47:26,075 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173102.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:47:43,687 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7218, 1.5060, 4.9110, 1.9350, 4.4127, 4.0809, 4.4665, 4.3068], + device='cuda:1'), covar=tensor([0.0490, 0.4617, 0.0448, 0.4019, 0.1031, 0.0907, 0.0517, 0.0585], + device='cuda:1'), in_proj_covar=tensor([0.0625, 0.0637, 0.0686, 0.0619, 0.0701, 0.0606, 0.0604, 0.0669], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:47:54,973 INFO [train.py:901] (1/4) Epoch 22, batch 3400, loss[loss=0.2191, simple_loss=0.3007, pruned_loss=0.06874, over 8490.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2875, pruned_loss=0.06221, over 1612220.17 frames. ], batch size: 26, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:48:12,778 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173168.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:48:21,529 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4399, 1.4572, 4.5987, 1.7609, 4.1137, 3.8321, 4.1946, 4.0462], + device='cuda:1'), covar=tensor([0.0507, 0.4502, 0.0423, 0.3693, 0.0899, 0.0920, 0.0515, 0.0544], + device='cuda:1'), in_proj_covar=tensor([0.0625, 0.0636, 0.0685, 0.0618, 0.0700, 0.0604, 0.0603, 0.0669], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:48:28,295 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.505e+02 2.494e+02 3.128e+02 3.771e+02 6.972e+02, threshold=6.255e+02, percent-clipped=4.0 +2023-02-07 03:48:28,968 INFO [train.py:901] (1/4) Epoch 22, batch 3450, loss[loss=0.2562, simple_loss=0.3376, pruned_loss=0.08744, over 8256.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2881, pruned_loss=0.06252, over 1614968.97 frames. ], batch size: 24, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:48:39,555 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173207.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:48:42,446 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173211.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:49:05,730 INFO [train.py:901] (1/4) Epoch 22, batch 3500, loss[loss=0.2333, simple_loss=0.3113, pruned_loss=0.07768, over 8321.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2867, pruned_loss=0.06147, over 1614621.05 frames. ], batch size: 26, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:49:24,783 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 03:49:38,939 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.642e+02 3.082e+02 3.788e+02 9.506e+02, threshold=6.164e+02, percent-clipped=4.0 +2023-02-07 03:49:39,651 INFO [train.py:901] (1/4) Epoch 22, batch 3550, loss[loss=0.233, simple_loss=0.3144, pruned_loss=0.07583, over 8535.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2874, pruned_loss=0.06156, over 1613327.31 frames. ], batch size: 28, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:49:50,460 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7771, 4.7755, 4.2239, 2.0397, 4.2276, 4.2588, 4.3014, 4.1688], + device='cuda:1'), covar=tensor([0.0594, 0.0438, 0.1030, 0.4517, 0.0866, 0.0922, 0.1195, 0.0696], + device='cuda:1'), in_proj_covar=tensor([0.0523, 0.0434, 0.0432, 0.0536, 0.0427, 0.0447, 0.0426, 0.0386], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:50:00,037 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173322.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:03,364 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173327.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:06,893 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173331.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:10,304 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5622, 1.3735, 1.6552, 1.3018, 0.9284, 1.4244, 1.4695, 1.1477], + device='cuda:1'), covar=tensor([0.0580, 0.1276, 0.1640, 0.1460, 0.0598, 0.1492, 0.0735, 0.0726], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0158, 0.0099, 0.0163, 0.0111, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 03:50:14,763 INFO [train.py:901] (1/4) Epoch 22, batch 3600, loss[loss=0.1958, simple_loss=0.2914, pruned_loss=0.05012, over 8524.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2875, pruned_loss=0.06156, over 1617586.03 frames. ], batch size: 28, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:50:25,001 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173356.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:26,365 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173358.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:43,490 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173383.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:48,590 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.398e+02 3.034e+02 4.459e+02 8.281e+02, threshold=6.068e+02, percent-clipped=7.0 +2023-02-07 03:50:49,318 INFO [train.py:901] (1/4) Epoch 22, batch 3650, loss[loss=0.1946, simple_loss=0.2822, pruned_loss=0.05347, over 8309.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2888, pruned_loss=0.06216, over 1618688.44 frames. ], batch size: 25, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:51:23,353 INFO [train.py:901] (1/4) Epoch 22, batch 3700, loss[loss=0.2609, simple_loss=0.331, pruned_loss=0.0954, over 7368.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2892, pruned_loss=0.06242, over 1618503.33 frames. ], batch size: 71, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:51:24,741 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 03:51:42,307 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173467.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:51:51,707 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-07 03:51:57,921 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.519e+02 2.931e+02 3.909e+02 7.363e+02, threshold=5.861e+02, percent-clipped=2.0 +2023-02-07 03:51:58,522 INFO [train.py:901] (1/4) Epoch 22, batch 3750, loss[loss=0.2662, simple_loss=0.3337, pruned_loss=0.09936, over 8358.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.289, pruned_loss=0.06287, over 1614953.20 frames. ], batch size: 26, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:51:58,726 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173492.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:52:11,076 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173509.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 03:52:12,776 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173512.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:52:18,135 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7209, 1.3985, 4.9318, 1.7796, 4.3983, 4.1320, 4.4959, 4.3433], + device='cuda:1'), covar=tensor([0.0556, 0.4679, 0.0407, 0.3862, 0.0959, 0.0903, 0.0532, 0.0579], + device='cuda:1'), in_proj_covar=tensor([0.0628, 0.0640, 0.0689, 0.0621, 0.0702, 0.0607, 0.0605, 0.0674], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:52:22,066 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9126, 1.4978, 1.7449, 1.3873, 0.9565, 1.4857, 1.7426, 1.5124], + device='cuda:1'), covar=tensor([0.0538, 0.1207, 0.1588, 0.1409, 0.0605, 0.1467, 0.0696, 0.0644], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0151, 0.0188, 0.0157, 0.0099, 0.0162, 0.0111, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 03:52:23,510 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 03:52:32,199 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1809, 1.5209, 1.7720, 1.4645, 1.0141, 1.5238, 1.9262, 1.7813], + device='cuda:1'), covar=tensor([0.0538, 0.1258, 0.1634, 0.1407, 0.0630, 0.1491, 0.0698, 0.0625], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0157, 0.0099, 0.0162, 0.0111, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 03:52:32,727 INFO [train.py:901] (1/4) Epoch 22, batch 3800, loss[loss=0.1883, simple_loss=0.2741, pruned_loss=0.0512, over 8623.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2899, pruned_loss=0.06347, over 1617574.69 frames. ], batch size: 39, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:52:35,962 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.34 vs. limit=5.0 +2023-02-07 03:52:41,038 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5686, 1.5594, 2.0944, 1.3911, 1.2464, 2.0987, 0.4079, 1.2124], + device='cuda:1'), covar=tensor([0.1579, 0.1263, 0.0336, 0.1061, 0.2552, 0.0367, 0.2118, 0.1286], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0198, 0.0127, 0.0220, 0.0267, 0.0135, 0.0170, 0.0192], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 03:52:58,746 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173578.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:53:07,973 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.750e+02 2.467e+02 3.140e+02 3.842e+02 8.904e+02, threshold=6.281e+02, percent-clipped=2.0 +2023-02-07 03:53:08,694 INFO [train.py:901] (1/4) Epoch 22, batch 3850, loss[loss=0.1826, simple_loss=0.255, pruned_loss=0.05511, over 7654.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2891, pruned_loss=0.06322, over 1609497.10 frames. ], batch size: 19, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:53:16,528 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173603.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:53:30,754 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 03:53:33,542 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173627.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:53:36,877 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173632.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:53:43,568 INFO [train.py:901] (1/4) Epoch 22, batch 3900, loss[loss=0.1947, simple_loss=0.2704, pruned_loss=0.0595, over 7675.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.289, pruned_loss=0.06313, over 1606536.05 frames. ], batch size: 18, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:54:02,783 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5190, 1.4832, 1.8604, 1.2704, 1.1921, 1.8637, 0.2618, 1.1664], + device='cuda:1'), covar=tensor([0.1720, 0.1260, 0.0431, 0.0952, 0.2916, 0.0483, 0.2241, 0.1276], + device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0198, 0.0128, 0.0221, 0.0269, 0.0136, 0.0171, 0.0193], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 03:54:03,324 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173671.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:54:17,229 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.938e+02 2.507e+02 2.945e+02 3.654e+02 8.206e+02, threshold=5.890e+02, percent-clipped=3.0 +2023-02-07 03:54:17,881 INFO [train.py:901] (1/4) Epoch 22, batch 3950, loss[loss=0.2063, simple_loss=0.2809, pruned_loss=0.06589, over 7819.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2892, pruned_loss=0.06351, over 1608145.05 frames. ], batch size: 20, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:54:53,286 INFO [train.py:901] (1/4) Epoch 22, batch 4000, loss[loss=0.2334, simple_loss=0.3157, pruned_loss=0.07552, over 8642.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2885, pruned_loss=0.06322, over 1607071.07 frames. ], batch size: 39, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:55:17,297 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.06 vs. limit=5.0 +2023-02-07 03:55:23,156 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173786.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:55:26,126 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.329e+02 2.821e+02 3.599e+02 1.045e+03, threshold=5.642e+02, percent-clipped=6.0 +2023-02-07 03:55:26,800 INFO [train.py:901] (1/4) Epoch 22, batch 4050, loss[loss=0.2133, simple_loss=0.2841, pruned_loss=0.07124, over 7809.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2885, pruned_loss=0.06284, over 1609348.14 frames. ], batch size: 20, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:56:02,760 INFO [train.py:901] (1/4) Epoch 22, batch 4100, loss[loss=0.211, simple_loss=0.2951, pruned_loss=0.06344, over 8561.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2882, pruned_loss=0.06299, over 1606409.74 frames. ], batch size: 49, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:56:10,366 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173853.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 03:56:23,169 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9216, 1.4831, 6.0909, 2.2893, 5.4759, 5.1007, 5.6534, 5.5084], + device='cuda:1'), covar=tensor([0.0496, 0.5002, 0.0320, 0.3684, 0.0987, 0.0913, 0.0531, 0.0555], + device='cuda:1'), in_proj_covar=tensor([0.0634, 0.0648, 0.0693, 0.0627, 0.0707, 0.0611, 0.0610, 0.0680], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:56:31,244 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173883.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:56:36,365 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.329e+02 2.764e+02 3.605e+02 7.317e+02, threshold=5.528e+02, percent-clipped=2.0 +2023-02-07 03:56:37,021 INFO [train.py:901] (1/4) Epoch 22, batch 4150, loss[loss=0.1829, simple_loss=0.2659, pruned_loss=0.04994, over 7658.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2874, pruned_loss=0.06231, over 1607644.06 frames. ], batch size: 19, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:56:47,601 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173908.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:57:12,186 INFO [train.py:901] (1/4) Epoch 22, batch 4200, loss[loss=0.2404, simple_loss=0.3216, pruned_loss=0.07959, over 8197.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2884, pruned_loss=0.06283, over 1609964.83 frames. ], batch size: 23, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:57:27,001 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8417, 6.0917, 5.1601, 2.6031, 5.2966, 5.6141, 5.5469, 5.4137], + device='cuda:1'), covar=tensor([0.0556, 0.0389, 0.0925, 0.4610, 0.0782, 0.1006, 0.1239, 0.0691], + device='cuda:1'), in_proj_covar=tensor([0.0525, 0.0435, 0.0432, 0.0536, 0.0423, 0.0445, 0.0426, 0.0387], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:57:28,899 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 03:57:29,744 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173968.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 03:57:35,169 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173976.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:57:46,760 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.784e+02 2.391e+02 3.056e+02 3.931e+02 9.713e+02, threshold=6.111e+02, percent-clipped=5.0 +2023-02-07 03:57:46,780 INFO [train.py:901] (1/4) Epoch 22, batch 4250, loss[loss=0.2355, simple_loss=0.3206, pruned_loss=0.07525, over 8511.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2881, pruned_loss=0.06242, over 1611880.03 frames. ], batch size: 26, lr: 3.44e-03, grad_scale: 4.0 +2023-02-07 03:57:55,806 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 03:58:04,801 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6323, 4.6188, 4.1842, 2.2061, 4.1510, 4.1445, 4.2655, 4.0431], + device='cuda:1'), covar=tensor([0.0626, 0.0498, 0.1001, 0.4270, 0.0782, 0.0983, 0.1193, 0.0752], + device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0436, 0.0433, 0.0538, 0.0425, 0.0448, 0.0428, 0.0388], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 03:58:15,723 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174033.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:58:17,063 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174035.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:58:21,635 INFO [train.py:901] (1/4) Epoch 22, batch 4300, loss[loss=0.2015, simple_loss=0.2936, pruned_loss=0.05467, over 8260.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2869, pruned_loss=0.0617, over 1613345.32 frames. ], batch size: 24, lr: 3.44e-03, grad_scale: 4.0 +2023-02-07 03:58:21,843 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174042.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:58:25,980 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 03:58:40,404 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174067.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:58:56,913 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174091.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:58:57,380 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.427e+02 2.775e+02 3.458e+02 5.995e+02, threshold=5.550e+02, percent-clipped=0.0 +2023-02-07 03:58:57,401 INFO [train.py:901] (1/4) Epoch 22, batch 4350, loss[loss=0.1863, simple_loss=0.2726, pruned_loss=0.04998, over 8567.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.287, pruned_loss=0.06173, over 1613811.16 frames. ], batch size: 31, lr: 3.43e-03, grad_scale: 4.0 +2023-02-07 03:59:25,055 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 03:59:32,552 INFO [train.py:901] (1/4) Epoch 22, batch 4400, loss[loss=0.1622, simple_loss=0.2382, pruned_loss=0.04306, over 7685.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2861, pruned_loss=0.06137, over 1611470.78 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 03:59:36,062 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7610, 2.9688, 2.5079, 4.0816, 1.7009, 2.2734, 2.5613, 3.3796], + device='cuda:1'), covar=tensor([0.0619, 0.0767, 0.0832, 0.0238, 0.1121, 0.1112, 0.0941, 0.0663], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0199, 0.0246, 0.0216, 0.0208, 0.0247, 0.0250, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 03:59:42,814 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174157.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:00:06,463 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 04:00:07,767 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.623e+02 3.065e+02 3.902e+02 1.119e+03, threshold=6.129e+02, percent-clipped=5.0 +2023-02-07 04:00:07,787 INFO [train.py:901] (1/4) Epoch 22, batch 4450, loss[loss=0.2123, simple_loss=0.294, pruned_loss=0.06529, over 8023.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2859, pruned_loss=0.0613, over 1610949.34 frames. ], batch size: 22, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:00:24,476 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-07 04:00:26,072 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5099, 1.8895, 2.9090, 1.4372, 2.2060, 1.9070, 1.6357, 2.0948], + device='cuda:1'), covar=tensor([0.1933, 0.2570, 0.0883, 0.4577, 0.1761, 0.3282, 0.2277, 0.2448], + device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0605, 0.0559, 0.0645, 0.0646, 0.0591, 0.0536, 0.0630], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:00:30,153 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174224.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 04:00:41,755 INFO [train.py:901] (1/4) Epoch 22, batch 4500, loss[loss=0.2141, simple_loss=0.2971, pruned_loss=0.0656, over 8479.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2864, pruned_loss=0.06149, over 1615705.01 frames. ], batch size: 28, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:00:46,572 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174249.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 04:00:56,995 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 04:01:17,050 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.524e+02 3.306e+02 4.354e+02 7.569e+02, threshold=6.612e+02, percent-clipped=6.0 +2023-02-07 04:01:17,070 INFO [train.py:901] (1/4) Epoch 22, batch 4550, loss[loss=0.243, simple_loss=0.3102, pruned_loss=0.08786, over 6971.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2877, pruned_loss=0.06222, over 1616949.70 frames. ], batch size: 73, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:01:23,308 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174301.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:01:28,573 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174309.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:01:44,046 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 +2023-02-07 04:01:51,062 INFO [train.py:901] (1/4) Epoch 22, batch 4600, loss[loss=0.1769, simple_loss=0.2589, pruned_loss=0.04744, over 7693.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.287, pruned_loss=0.06167, over 1615342.65 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:01:54,750 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174347.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:02:12,033 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174372.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:02:15,443 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174377.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:02:16,890 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174379.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:02:25,988 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.714e+02 2.375e+02 2.973e+02 3.873e+02 1.031e+03, threshold=5.946e+02, percent-clipped=3.0 +2023-02-07 04:02:26,008 INFO [train.py:901] (1/4) Epoch 22, batch 4650, loss[loss=0.1662, simple_loss=0.2525, pruned_loss=0.03991, over 7921.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2864, pruned_loss=0.06135, over 1614660.38 frames. ], batch size: 20, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:02:31,945 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 04:02:37,942 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174406.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:03:02,512 INFO [train.py:901] (1/4) Epoch 22, batch 4700, loss[loss=0.2367, simple_loss=0.3284, pruned_loss=0.07251, over 8334.00 frames. ], tot_loss[loss=0.204, simple_loss=0.286, pruned_loss=0.06097, over 1608879.52 frames. ], batch size: 26, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:03:21,332 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6299, 1.8415, 1.9478, 1.3238, 2.0407, 1.3903, 0.5887, 1.8491], + device='cuda:1'), covar=tensor([0.0597, 0.0368, 0.0325, 0.0565, 0.0422, 0.0858, 0.0887, 0.0317], + device='cuda:1'), in_proj_covar=tensor([0.0449, 0.0387, 0.0343, 0.0442, 0.0372, 0.0528, 0.0385, 0.0415], + device='cuda:1'), out_proj_covar=tensor([1.2022e-04, 1.0130e-04, 9.0373e-05, 1.1635e-04, 9.7755e-05, 1.4885e-04, + 1.0387e-04, 1.0982e-04], device='cuda:1') +2023-02-07 04:03:37,053 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.374e+02 2.923e+02 3.899e+02 9.329e+02, threshold=5.846e+02, percent-clipped=2.0 +2023-02-07 04:03:37,073 INFO [train.py:901] (1/4) Epoch 22, batch 4750, loss[loss=0.2009, simple_loss=0.287, pruned_loss=0.05744, over 7821.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2855, pruned_loss=0.06043, over 1608633.22 frames. ], batch size: 20, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:03:37,263 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174492.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:03:38,614 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174494.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:03:43,328 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174501.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:03:59,886 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1865, 1.0512, 1.2618, 1.0113, 1.0073, 1.2930, 0.1087, 0.8851], + device='cuda:1'), covar=tensor([0.1679, 0.1507, 0.0505, 0.0888, 0.2884, 0.0593, 0.2222, 0.1348], + device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0198, 0.0128, 0.0220, 0.0269, 0.0136, 0.0171, 0.0194], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 04:04:04,468 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 04:04:06,505 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 04:04:12,381 INFO [train.py:901] (1/4) Epoch 22, batch 4800, loss[loss=0.2098, simple_loss=0.3014, pruned_loss=0.0591, over 8725.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2866, pruned_loss=0.06087, over 1605226.01 frames. ], batch size: 30, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:04:19,970 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.71 vs. limit=5.0 +2023-02-07 04:04:22,158 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7902, 1.7172, 2.5063, 1.5153, 1.2799, 2.4682, 0.5345, 1.4308], + device='cuda:1'), covar=tensor([0.1790, 0.1404, 0.0340, 0.1441, 0.2962, 0.0396, 0.2398, 0.1557], + device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0198, 0.0129, 0.0221, 0.0270, 0.0137, 0.0172, 0.0195], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 04:04:27,165 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 04:04:46,095 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.398e+02 2.995e+02 3.860e+02 8.125e+02, threshold=5.990e+02, percent-clipped=3.0 +2023-02-07 04:04:46,115 INFO [train.py:901] (1/4) Epoch 22, batch 4850, loss[loss=0.1996, simple_loss=0.2742, pruned_loss=0.06251, over 7687.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2872, pruned_loss=0.06153, over 1606993.83 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:04:55,435 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 04:05:02,312 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174616.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:05:20,309 INFO [train.py:901] (1/4) Epoch 22, batch 4900, loss[loss=0.1791, simple_loss=0.2546, pruned_loss=0.05183, over 7430.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2871, pruned_loss=0.06139, over 1604715.01 frames. ], batch size: 17, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:05:23,130 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174645.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:05:29,309 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174653.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:05:56,491 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.582e+02 3.121e+02 3.821e+02 7.682e+02, threshold=6.242e+02, percent-clipped=2.0 +2023-02-07 04:05:56,518 INFO [train.py:901] (1/4) Epoch 22, batch 4950, loss[loss=0.1707, simple_loss=0.2499, pruned_loss=0.04582, over 7700.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2875, pruned_loss=0.0617, over 1601395.31 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:06:17,805 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:21,291 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3858, 2.2397, 1.7337, 2.1235, 1.8639, 1.2939, 1.8182, 1.9315], + device='cuda:1'), covar=tensor([0.1476, 0.0441, 0.1355, 0.0562, 0.0839, 0.1825, 0.1069, 0.0879], + device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0231, 0.0331, 0.0305, 0.0297, 0.0336, 0.0339, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 04:06:27,439 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3669, 1.4898, 1.3880, 1.7914, 0.7132, 1.2633, 1.3047, 1.4602], + device='cuda:1'), covar=tensor([0.0851, 0.0724, 0.0951, 0.0484, 0.1109, 0.1234, 0.0706, 0.0714], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0198, 0.0245, 0.0216, 0.0208, 0.0248, 0.0250, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 04:06:30,666 INFO [train.py:901] (1/4) Epoch 22, batch 5000, loss[loss=0.1954, simple_loss=0.2808, pruned_loss=0.05506, over 8035.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2877, pruned_loss=0.06128, over 1609387.40 frames. ], batch size: 22, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:06:35,039 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174748.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:36,260 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174750.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:36,425 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174750.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:44,706 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174760.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:51,033 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:54,597 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174773.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:56,004 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174775.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:07:07,684 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.477e+02 2.928e+02 3.612e+02 7.754e+02, threshold=5.856e+02, percent-clipped=3.0 +2023-02-07 04:07:07,711 INFO [train.py:901] (1/4) Epoch 22, batch 5050, loss[loss=0.1695, simple_loss=0.2537, pruned_loss=0.04263, over 7531.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2876, pruned_loss=0.06133, over 1606719.71 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:07:16,964 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1571, 1.3263, 1.6109, 1.2363, 0.7393, 1.4082, 1.2513, 1.0188], + device='cuda:1'), covar=tensor([0.0598, 0.1271, 0.1570, 0.1453, 0.0576, 0.1463, 0.0674, 0.0725], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0159, 0.0100, 0.0164, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 04:07:27,839 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4149, 2.3223, 3.1683, 2.6031, 3.0394, 2.5413, 2.2045, 1.9115], + device='cuda:1'), covar=tensor([0.5511, 0.5080, 0.2010, 0.3759, 0.2467, 0.2754, 0.1852, 0.5359], + device='cuda:1'), in_proj_covar=tensor([0.0935, 0.0975, 0.0800, 0.0941, 0.0989, 0.0888, 0.0745, 0.0821], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 04:07:34,470 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 04:07:42,661 INFO [train.py:901] (1/4) Epoch 22, batch 5100, loss[loss=0.2317, simple_loss=0.316, pruned_loss=0.07372, over 8293.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.288, pruned_loss=0.06117, over 1612665.67 frames. ], batch size: 23, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:07:58,167 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:08:03,624 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174872.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:08:17,820 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.498e+02 3.130e+02 3.757e+02 7.363e+02, threshold=6.259e+02, percent-clipped=3.0 +2023-02-07 04:08:17,841 INFO [train.py:901] (1/4) Epoch 22, batch 5150, loss[loss=0.2295, simple_loss=0.3037, pruned_loss=0.0777, over 8505.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2875, pruned_loss=0.06106, over 1614380.83 frames. ], batch size: 28, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:08:21,142 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174897.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:08:52,070 INFO [train.py:901] (1/4) Epoch 22, batch 5200, loss[loss=0.2027, simple_loss=0.2777, pruned_loss=0.06386, over 6781.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2884, pruned_loss=0.06195, over 1613613.68 frames. ], batch size: 71, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:08:52,925 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2538, 1.7068, 4.3934, 2.0719, 3.9381, 3.6974, 4.0368, 3.8979], + device='cuda:1'), covar=tensor([0.0626, 0.4578, 0.0567, 0.3821, 0.1089, 0.1009, 0.0600, 0.0687], + device='cuda:1'), in_proj_covar=tensor([0.0628, 0.0646, 0.0695, 0.0627, 0.0708, 0.0603, 0.0606, 0.0677], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:08:58,259 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2182, 4.1628, 3.7656, 1.9057, 3.6859, 3.7931, 3.8068, 3.6190], + device='cuda:1'), covar=tensor([0.0747, 0.0571, 0.1081, 0.5008, 0.0926, 0.1095, 0.1221, 0.0907], + device='cuda:1'), in_proj_covar=tensor([0.0522, 0.0433, 0.0432, 0.0536, 0.0424, 0.0444, 0.0424, 0.0387], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:09:26,985 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 2.432e+02 2.854e+02 3.739e+02 7.258e+02, threshold=5.708e+02, percent-clipped=1.0 +2023-02-07 04:09:27,005 INFO [train.py:901] (1/4) Epoch 22, batch 5250, loss[loss=0.1901, simple_loss=0.273, pruned_loss=0.05361, over 7421.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2883, pruned_loss=0.06186, over 1609480.70 frames. ], batch size: 17, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:09:31,814 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 04:09:44,046 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175016.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:09:49,542 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175024.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:10:01,589 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175041.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:10:02,004 INFO [train.py:901] (1/4) Epoch 22, batch 5300, loss[loss=0.2042, simple_loss=0.2849, pruned_loss=0.06174, over 8026.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2886, pruned_loss=0.06252, over 1612653.02 frames. ], batch size: 22, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:10:07,099 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175049.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:10:19,119 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175067.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:10:35,763 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.519e+02 3.149e+02 3.909e+02 1.075e+03, threshold=6.297e+02, percent-clipped=6.0 +2023-02-07 04:10:35,783 INFO [train.py:901] (1/4) Epoch 22, batch 5350, loss[loss=0.1775, simple_loss=0.2552, pruned_loss=0.04993, over 7540.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2883, pruned_loss=0.06243, over 1607328.12 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:10:57,900 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175121.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:11:11,832 INFO [train.py:901] (1/4) Epoch 22, batch 5400, loss[loss=0.2056, simple_loss=0.2877, pruned_loss=0.06172, over 8470.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2883, pruned_loss=0.06258, over 1611015.93 frames. ], batch size: 25, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:11:13,259 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8171, 3.7595, 3.4029, 1.9076, 3.3089, 3.5101, 3.3924, 3.4128], + device='cuda:1'), covar=tensor([0.0870, 0.0676, 0.1130, 0.4773, 0.0936, 0.1141, 0.1389, 0.0842], + device='cuda:1'), in_proj_covar=tensor([0.0521, 0.0432, 0.0430, 0.0533, 0.0422, 0.0442, 0.0422, 0.0384], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:11:14,644 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175146.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:11:22,727 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175157.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:11:39,695 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175182.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:11:46,267 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.426e+02 2.833e+02 4.034e+02 1.686e+03, threshold=5.665e+02, percent-clipped=5.0 +2023-02-07 04:11:46,288 INFO [train.py:901] (1/4) Epoch 22, batch 5450, loss[loss=0.1884, simple_loss=0.2615, pruned_loss=0.05761, over 8240.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2878, pruned_loss=0.06201, over 1610720.26 frames. ], batch size: 22, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:12:09,778 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175225.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:12:20,249 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 04:12:22,321 INFO [train.py:901] (1/4) Epoch 22, batch 5500, loss[loss=0.1766, simple_loss=0.2594, pruned_loss=0.04693, over 5124.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2877, pruned_loss=0.06166, over 1607476.38 frames. ], batch size: 11, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:12:22,568 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4480, 1.6024, 2.1039, 1.3692, 1.4756, 1.7087, 1.4658, 1.4556], + device='cuda:1'), covar=tensor([0.1822, 0.2295, 0.0997, 0.4319, 0.1848, 0.3134, 0.2289, 0.2046], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0605, 0.0555, 0.0643, 0.0646, 0.0591, 0.0535, 0.0629], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:12:40,056 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-07 04:12:56,620 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.376e+02 2.848e+02 3.508e+02 8.289e+02, threshold=5.697e+02, percent-clipped=6.0 +2023-02-07 04:12:56,641 INFO [train.py:901] (1/4) Epoch 22, batch 5550, loss[loss=0.1885, simple_loss=0.261, pruned_loss=0.05801, over 7438.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2874, pruned_loss=0.06204, over 1606403.27 frames. ], batch size: 17, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:13:31,799 INFO [train.py:901] (1/4) Epoch 22, batch 5600, loss[loss=0.1823, simple_loss=0.2661, pruned_loss=0.0492, over 8102.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2887, pruned_loss=0.06195, over 1610637.81 frames. ], batch size: 21, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:13:46,074 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8480, 2.2693, 3.7917, 1.6856, 2.8733, 2.3229, 1.8558, 2.8958], + device='cuda:1'), covar=tensor([0.1812, 0.2422, 0.0890, 0.4209, 0.1799, 0.2932, 0.2279, 0.2209], + device='cuda:1'), in_proj_covar=tensor([0.0525, 0.0604, 0.0554, 0.0641, 0.0645, 0.0589, 0.0535, 0.0629], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:13:49,361 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2003, 1.0936, 1.3040, 1.0826, 1.0469, 1.3271, 0.0750, 0.9600], + device='cuda:1'), covar=tensor([0.1555, 0.1276, 0.0473, 0.0665, 0.2510, 0.0541, 0.1960, 0.1206], + device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0197, 0.0129, 0.0220, 0.0267, 0.0137, 0.0168, 0.0194], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 04:13:50,743 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.40 vs. limit=5.0 +2023-02-07 04:14:03,194 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4669, 1.5156, 1.4271, 1.8404, 0.7007, 1.3385, 1.2879, 1.5269], + device='cuda:1'), covar=tensor([0.0831, 0.0751, 0.1044, 0.0544, 0.1132, 0.1288, 0.0799, 0.0679], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0198, 0.0247, 0.0216, 0.0208, 0.0247, 0.0252, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 04:14:03,828 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175388.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:14:06,409 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.734e+02 2.532e+02 3.141e+02 4.135e+02 1.836e+03, threshold=6.283e+02, percent-clipped=10.0 +2023-02-07 04:14:06,429 INFO [train.py:901] (1/4) Epoch 22, batch 5650, loss[loss=0.17, simple_loss=0.2562, pruned_loss=0.04186, over 8237.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2874, pruned_loss=0.0616, over 1609302.10 frames. ], batch size: 22, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:14:09,947 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7032, 2.1504, 3.2485, 1.5722, 2.6243, 2.0483, 1.8060, 2.5843], + device='cuda:1'), covar=tensor([0.1953, 0.2565, 0.0889, 0.4452, 0.1778, 0.3264, 0.2435, 0.2248], + device='cuda:1'), in_proj_covar=tensor([0.0525, 0.0604, 0.0553, 0.0640, 0.0645, 0.0589, 0.0535, 0.0628], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:14:20,099 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175412.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:14:22,665 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 04:14:37,557 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175438.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:14:40,743 INFO [train.py:901] (1/4) Epoch 22, batch 5700, loss[loss=0.202, simple_loss=0.2817, pruned_loss=0.06115, over 8026.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2875, pruned_loss=0.06201, over 1610693.46 frames. ], batch size: 22, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:14:56,158 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175463.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:15:15,528 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.452e+02 2.878e+02 3.661e+02 5.836e+02, threshold=5.755e+02, percent-clipped=0.0 +2023-02-07 04:15:15,548 INFO [train.py:901] (1/4) Epoch 22, batch 5750, loss[loss=0.2151, simple_loss=0.3036, pruned_loss=0.06331, over 8187.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2879, pruned_loss=0.06186, over 1615317.89 frames. ], batch size: 23, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:15:20,411 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175499.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:15:22,436 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175501.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:15:27,232 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 04:15:50,252 INFO [train.py:901] (1/4) Epoch 22, batch 5800, loss[loss=0.1758, simple_loss=0.2464, pruned_loss=0.05259, over 7437.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2881, pruned_loss=0.06231, over 1612904.69 frames. ], batch size: 17, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:16:09,022 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175569.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:16:25,894 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 2.280e+02 2.739e+02 3.457e+02 6.413e+02, threshold=5.479e+02, percent-clipped=3.0 +2023-02-07 04:16:25,914 INFO [train.py:901] (1/4) Epoch 22, batch 5850, loss[loss=0.2113, simple_loss=0.3018, pruned_loss=0.0604, over 8516.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2873, pruned_loss=0.06182, over 1612236.50 frames. ], batch size: 28, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:16:42,238 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175616.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:17:00,352 INFO [train.py:901] (1/4) Epoch 22, batch 5900, loss[loss=0.1672, simple_loss=0.2526, pruned_loss=0.04091, over 7801.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2882, pruned_loss=0.06217, over 1611737.13 frames. ], batch size: 20, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:17:29,304 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175684.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:17:35,263 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.520e+02 3.043e+02 3.699e+02 9.671e+02, threshold=6.086e+02, percent-clipped=7.0 +2023-02-07 04:17:35,284 INFO [train.py:901] (1/4) Epoch 22, batch 5950, loss[loss=0.2035, simple_loss=0.2925, pruned_loss=0.05725, over 8293.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2878, pruned_loss=0.06184, over 1611072.31 frames. ], batch size: 23, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:18:00,062 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175728.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:18:02,743 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175732.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:18:09,573 INFO [train.py:901] (1/4) Epoch 22, batch 6000, loss[loss=0.1963, simple_loss=0.2784, pruned_loss=0.05711, over 7656.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2873, pruned_loss=0.06164, over 1609779.45 frames. ], batch size: 19, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:18:09,573 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 04:18:21,635 INFO [train.py:935] (1/4) Epoch 22, validation: loss=0.1729, simple_loss=0.2732, pruned_loss=0.03632, over 944034.00 frames. +2023-02-07 04:18:21,635 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 04:18:28,878 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2364, 2.0950, 2.7489, 2.3454, 2.7478, 2.2534, 2.1088, 1.6716], + device='cuda:1'), covar=tensor([0.5538, 0.4931, 0.1997, 0.3506, 0.2344, 0.3104, 0.1919, 0.5268], + device='cuda:1'), in_proj_covar=tensor([0.0937, 0.0977, 0.0802, 0.0942, 0.0992, 0.0892, 0.0745, 0.0824], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 04:18:31,484 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175756.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:18:36,998 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9608, 2.0997, 1.7263, 2.6971, 1.1833, 1.6114, 1.8194, 2.0695], + device='cuda:1'), covar=tensor([0.0781, 0.0795, 0.0956, 0.0353, 0.1129, 0.1251, 0.0862, 0.0805], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0197, 0.0246, 0.0216, 0.0207, 0.0246, 0.0251, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 04:18:56,217 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.478e+02 2.934e+02 3.623e+02 7.032e+02, threshold=5.869e+02, percent-clipped=2.0 +2023-02-07 04:18:56,238 INFO [train.py:901] (1/4) Epoch 22, batch 6050, loss[loss=0.2237, simple_loss=0.3123, pruned_loss=0.06756, over 8474.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2875, pruned_loss=0.06193, over 1612184.44 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:19:06,541 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0046, 2.2716, 1.7612, 2.7845, 1.3042, 1.5902, 1.9167, 2.2405], + device='cuda:1'), covar=tensor([0.0710, 0.0703, 0.0919, 0.0358, 0.1094, 0.1275, 0.0887, 0.0699], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0197, 0.0246, 0.0216, 0.0206, 0.0246, 0.0251, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 04:19:31,868 INFO [train.py:901] (1/4) Epoch 22, batch 6100, loss[loss=0.212, simple_loss=0.2853, pruned_loss=0.06936, over 6848.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2877, pruned_loss=0.06187, over 1614834.71 frames. ], batch size: 71, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:19:32,675 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:19:35,616 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175847.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:19:51,982 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175871.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:19:52,717 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175872.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:19:56,553 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 04:20:07,199 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.365e+02 2.974e+02 3.880e+02 6.577e+02, threshold=5.949e+02, percent-clipped=2.0 +2023-02-07 04:20:07,219 INFO [train.py:901] (1/4) Epoch 22, batch 6150, loss[loss=0.2673, simple_loss=0.3391, pruned_loss=0.09778, over 8558.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2862, pruned_loss=0.06083, over 1614275.78 frames. ], batch size: 31, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:20:10,661 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175897.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:11,996 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175899.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:20,145 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2152, 2.0128, 2.6343, 2.1814, 2.5890, 2.2654, 2.0447, 1.3074], + device='cuda:1'), covar=tensor([0.5396, 0.4650, 0.1924, 0.3677, 0.2441, 0.2959, 0.1921, 0.5313], + device='cuda:1'), in_proj_covar=tensor([0.0943, 0.0981, 0.0806, 0.0946, 0.0997, 0.0896, 0.0748, 0.0826], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 04:20:40,362 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175940.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:41,485 INFO [train.py:901] (1/4) Epoch 22, batch 6200, loss[loss=0.226, simple_loss=0.3101, pruned_loss=0.07094, over 8448.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2856, pruned_loss=0.06017, over 1613323.27 frames. ], batch size: 27, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:20:52,906 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175958.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:56,910 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175963.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:58,222 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175965.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:21:06,753 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3639, 2.2854, 1.6656, 2.1065, 1.9719, 1.4056, 1.7911, 1.7978], + device='cuda:1'), covar=tensor([0.1630, 0.0440, 0.1403, 0.0645, 0.0747, 0.1777, 0.1112, 0.1049], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0233, 0.0335, 0.0310, 0.0299, 0.0340, 0.0345, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 04:21:15,650 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.329e+02 2.882e+02 3.634e+02 1.217e+03, threshold=5.765e+02, percent-clipped=6.0 +2023-02-07 04:21:15,671 INFO [train.py:901] (1/4) Epoch 22, batch 6250, loss[loss=0.2205, simple_loss=0.3064, pruned_loss=0.06729, over 8245.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2865, pruned_loss=0.06106, over 1617261.67 frames. ], batch size: 24, lr: 3.42e-03, grad_scale: 16.0 +2023-02-07 04:21:51,347 INFO [train.py:901] (1/4) Epoch 22, batch 6300, loss[loss=0.2121, simple_loss=0.2901, pruned_loss=0.06703, over 7792.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2858, pruned_loss=0.06064, over 1617176.25 frames. ], batch size: 20, lr: 3.42e-03, grad_scale: 16.0 +2023-02-07 04:22:12,730 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176072.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:22:26,686 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.297e+02 2.795e+02 3.577e+02 6.374e+02, threshold=5.590e+02, percent-clipped=1.0 +2023-02-07 04:22:26,706 INFO [train.py:901] (1/4) Epoch 22, batch 6350, loss[loss=0.1994, simple_loss=0.2864, pruned_loss=0.05622, over 7975.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2856, pruned_loss=0.06015, over 1615571.03 frames. ], batch size: 21, lr: 3.42e-03, grad_scale: 16.0 +2023-02-07 04:22:34,421 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176103.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:22:51,215 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176127.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:22:51,826 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176128.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:23:01,405 INFO [train.py:901] (1/4) Epoch 22, batch 6400, loss[loss=0.183, simple_loss=0.2846, pruned_loss=0.04073, over 8641.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2867, pruned_loss=0.06079, over 1616034.85 frames. ], batch size: 34, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:23:08,435 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176152.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:23:33,468 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176187.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:23:36,652 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.375e+02 2.869e+02 3.334e+02 7.002e+02, threshold=5.738e+02, percent-clipped=1.0 +2023-02-07 04:23:36,672 INFO [train.py:901] (1/4) Epoch 22, batch 6450, loss[loss=0.1735, simple_loss=0.2558, pruned_loss=0.04559, over 7804.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2863, pruned_loss=0.06054, over 1613449.13 frames. ], batch size: 20, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:23:52,615 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:23:54,745 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 +2023-02-07 04:24:05,442 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-07 04:24:09,913 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:24:11,687 INFO [train.py:901] (1/4) Epoch 22, batch 6500, loss[loss=0.2244, simple_loss=0.3066, pruned_loss=0.07108, over 8369.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2864, pruned_loss=0.0602, over 1619742.03 frames. ], batch size: 24, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:24:12,449 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176243.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:24:23,888 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176260.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:24:27,371 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2682, 2.1145, 1.7778, 1.9178, 1.6520, 1.4153, 1.6067, 1.6683], + device='cuda:1'), covar=tensor([0.1238, 0.0399, 0.1223, 0.0537, 0.0768, 0.1547, 0.0979, 0.0810], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0234, 0.0335, 0.0311, 0.0300, 0.0342, 0.0347, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 04:24:32,064 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5074, 2.4369, 1.7493, 2.2071, 1.9834, 1.4728, 1.8944, 2.0270], + device='cuda:1'), covar=tensor([0.1524, 0.0438, 0.1359, 0.0612, 0.0743, 0.1660, 0.1061, 0.1039], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0233, 0.0334, 0.0311, 0.0300, 0.0341, 0.0347, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 04:24:34,066 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176275.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:24:45,633 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.325e+02 2.725e+02 3.404e+02 5.159e+02, threshold=5.450e+02, percent-clipped=0.0 +2023-02-07 04:24:45,652 INFO [train.py:901] (1/4) Epoch 22, batch 6550, loss[loss=0.2385, simple_loss=0.2991, pruned_loss=0.08896, over 7973.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.286, pruned_loss=0.05967, over 1620382.40 frames. ], batch size: 21, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:24:55,357 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6592, 2.1737, 4.2404, 1.5490, 3.2394, 2.2615, 1.6955, 3.2008], + device='cuda:1'), covar=tensor([0.1799, 0.2618, 0.0765, 0.4281, 0.1545, 0.2937, 0.2308, 0.1906], + device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0608, 0.0558, 0.0646, 0.0650, 0.0596, 0.0540, 0.0635], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:24:57,260 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176307.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:25:04,188 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2909, 1.9867, 2.5498, 2.0728, 2.4402, 2.2609, 2.1388, 1.3203], + device='cuda:1'), covar=tensor([0.5233, 0.4818, 0.1959, 0.3734, 0.2609, 0.2985, 0.1798, 0.5366], + device='cuda:1'), in_proj_covar=tensor([0.0941, 0.0979, 0.0806, 0.0945, 0.0997, 0.0896, 0.0749, 0.0828], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 04:25:09,353 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 04:25:21,083 INFO [train.py:901] (1/4) Epoch 22, batch 6600, loss[loss=0.2032, simple_loss=0.2781, pruned_loss=0.06416, over 7788.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2868, pruned_loss=0.0604, over 1617903.44 frames. ], batch size: 19, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:25:29,274 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 04:25:32,749 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176358.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:25:55,381 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.481e+02 2.510e+02 3.110e+02 4.060e+02 7.968e+02, threshold=6.221e+02, percent-clipped=4.0 +2023-02-07 04:25:55,402 INFO [train.py:901] (1/4) Epoch 22, batch 6650, loss[loss=0.2157, simple_loss=0.3057, pruned_loss=0.0629, over 8619.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2869, pruned_loss=0.06029, over 1619122.67 frames. ], batch size: 34, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:25:56,917 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6986, 2.2817, 1.8132, 4.0275, 1.6398, 1.5786, 2.2934, 2.7202], + device='cuda:1'), covar=tensor([0.1597, 0.1272, 0.1923, 0.0282, 0.1367, 0.1814, 0.1246, 0.0929], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0196, 0.0243, 0.0214, 0.0205, 0.0245, 0.0249, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 04:26:17,167 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176422.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:26:31,230 INFO [train.py:901] (1/4) Epoch 22, batch 6700, loss[loss=0.2127, simple_loss=0.2965, pruned_loss=0.06444, over 8441.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2864, pruned_loss=0.06009, over 1616460.94 frames. ], batch size: 27, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:26:32,111 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176443.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:26:49,612 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176468.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:27:00,364 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176484.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:27:05,580 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.672e+02 3.290e+02 4.002e+02 8.131e+02, threshold=6.579e+02, percent-clipped=6.0 +2023-02-07 04:27:05,600 INFO [train.py:901] (1/4) Epoch 22, batch 6750, loss[loss=0.1996, simple_loss=0.2802, pruned_loss=0.05951, over 8090.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2855, pruned_loss=0.05984, over 1613863.38 frames. ], batch size: 21, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:27:41,484 INFO [train.py:901] (1/4) Epoch 22, batch 6800, loss[loss=0.2229, simple_loss=0.3006, pruned_loss=0.07263, over 8464.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2869, pruned_loss=0.06087, over 1613960.20 frames. ], batch size: 27, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:27:44,996 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 04:28:10,349 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.46 vs. limit=2.0 +2023-02-07 04:28:16,782 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.317e+02 3.026e+02 3.783e+02 8.757e+02, threshold=6.052e+02, percent-clipped=1.0 +2023-02-07 04:28:16,802 INFO [train.py:901] (1/4) Epoch 22, batch 6850, loss[loss=0.1798, simple_loss=0.2499, pruned_loss=0.05488, over 7549.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2867, pruned_loss=0.06082, over 1615211.67 frames. ], batch size: 18, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:28:24,732 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176604.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:28:31,628 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176614.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:28:34,697 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 04:28:34,752 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176619.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:28:48,422 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176639.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:28:50,268 INFO [train.py:901] (1/4) Epoch 22, batch 6900, loss[loss=0.2337, simple_loss=0.318, pruned_loss=0.07468, over 8458.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2872, pruned_loss=0.06133, over 1614267.74 frames. ], batch size: 27, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:28:51,045 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:17,275 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176678.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:20,584 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176683.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:26,754 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 2.439e+02 3.078e+02 3.806e+02 5.995e+02, threshold=6.157e+02, percent-clipped=0.0 +2023-02-07 04:29:26,774 INFO [train.py:901] (1/4) Epoch 22, batch 6950, loss[loss=0.1693, simple_loss=0.2513, pruned_loss=0.04359, over 8085.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2865, pruned_loss=0.0607, over 1614177.50 frames. ], batch size: 21, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:29:35,488 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176703.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:44,270 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 04:29:46,552 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176719.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:48,145 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 +2023-02-07 04:29:56,849 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176734.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:30:02,057 INFO [train.py:901] (1/4) Epoch 22, batch 7000, loss[loss=0.1515, simple_loss=0.2345, pruned_loss=0.03419, over 7257.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2864, pruned_loss=0.06092, over 1609914.51 frames. ], batch size: 16, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:30:03,961 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-07 04:30:37,821 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.601e+02 2.400e+02 2.923e+02 3.703e+02 8.900e+02, threshold=5.847e+02, percent-clipped=5.0 +2023-02-07 04:30:37,842 INFO [train.py:901] (1/4) Epoch 22, batch 7050, loss[loss=0.211, simple_loss=0.2897, pruned_loss=0.06612, over 8727.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2864, pruned_loss=0.06065, over 1614336.27 frames. ], batch size: 30, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:31:03,290 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176828.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:31:11,273 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0831, 1.5785, 1.3998, 1.4950, 1.3312, 1.2096, 1.3041, 1.2694], + device='cuda:1'), covar=tensor([0.1228, 0.0473, 0.1326, 0.0608, 0.0835, 0.1628, 0.0964, 0.0864], + device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0233, 0.0337, 0.0312, 0.0300, 0.0344, 0.0347, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 04:31:12,380 INFO [train.py:901] (1/4) Epoch 22, batch 7100, loss[loss=0.1681, simple_loss=0.2529, pruned_loss=0.04162, over 7912.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2868, pruned_loss=0.06096, over 1615719.65 frames. ], batch size: 20, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:31:31,798 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176871.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:31:46,090 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.664e+02 2.316e+02 2.836e+02 3.633e+02 7.093e+02, threshold=5.673e+02, percent-clipped=3.0 +2023-02-07 04:31:46,110 INFO [train.py:901] (1/4) Epoch 22, batch 7150, loss[loss=0.2845, simple_loss=0.3536, pruned_loss=0.1077, over 8458.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2871, pruned_loss=0.06121, over 1615008.57 frames. ], batch size: 27, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:31:51,637 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-07 04:32:22,264 INFO [train.py:901] (1/4) Epoch 22, batch 7200, loss[loss=0.2277, simple_loss=0.3255, pruned_loss=0.06495, over 8189.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.289, pruned_loss=0.06218, over 1613561.94 frames. ], batch size: 23, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:32:23,131 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176943.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:32:44,910 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176975.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:32:52,841 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176987.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:32:54,984 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176990.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:32:56,137 INFO [train.py:901] (1/4) Epoch 22, batch 7250, loss[loss=0.2676, simple_loss=0.3365, pruned_loss=0.09934, over 7267.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2889, pruned_loss=0.0622, over 1614413.16 frames. ], batch size: 71, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:32:56,787 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.385e+02 2.852e+02 3.441e+02 7.839e+02, threshold=5.703e+02, percent-clipped=2.0 +2023-02-07 04:33:02,404 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177000.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:33:10,522 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1849, 4.1175, 3.7868, 2.1372, 3.6567, 3.7961, 3.7087, 3.6371], + device='cuda:1'), covar=tensor([0.0747, 0.0577, 0.1054, 0.4497, 0.0933, 0.1085, 0.1383, 0.0893], + device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0438, 0.0432, 0.0542, 0.0427, 0.0448, 0.0428, 0.0388], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:33:14,104 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177015.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:33:15,422 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8713, 1.5305, 3.5626, 1.7058, 2.4999, 3.8536, 3.9244, 3.3470], + device='cuda:1'), covar=tensor([0.1199, 0.1742, 0.0249, 0.1833, 0.0947, 0.0199, 0.0437, 0.0497], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0319, 0.0284, 0.0313, 0.0309, 0.0265, 0.0418, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 04:33:21,899 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177027.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:33:31,982 INFO [train.py:901] (1/4) Epoch 22, batch 7300, loss[loss=0.1879, simple_loss=0.2666, pruned_loss=0.05456, over 7520.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2884, pruned_loss=0.06194, over 1611207.15 frames. ], batch size: 18, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:34:06,489 INFO [train.py:901] (1/4) Epoch 22, batch 7350, loss[loss=0.2257, simple_loss=0.2995, pruned_loss=0.07592, over 7922.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2886, pruned_loss=0.06226, over 1605444.44 frames. ], batch size: 20, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:34:07,156 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.532e+02 3.310e+02 4.342e+02 9.656e+02, threshold=6.621e+02, percent-clipped=7.0 +2023-02-07 04:34:13,468 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177102.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:34:26,059 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 04:34:31,743 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177127.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:34:33,081 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177129.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:34:42,503 INFO [train.py:901] (1/4) Epoch 22, batch 7400, loss[loss=0.1921, simple_loss=0.2894, pruned_loss=0.04745, over 8471.00 frames. ], tot_loss[loss=0.206, simple_loss=0.288, pruned_loss=0.06206, over 1609140.00 frames. ], batch size: 25, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:34:42,680 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177142.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:34:42,692 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7337, 1.9338, 1.6882, 2.3682, 1.0882, 1.5172, 1.7681, 1.8850], + device='cuda:1'), covar=tensor([0.0832, 0.0721, 0.0923, 0.0412, 0.1036, 0.1215, 0.0709, 0.0769], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0195, 0.0243, 0.0212, 0.0206, 0.0245, 0.0248, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 04:34:47,998 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 04:35:09,031 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1541, 3.0428, 2.8813, 1.6230, 2.7654, 2.8429, 2.8017, 2.7980], + device='cuda:1'), covar=tensor([0.1237, 0.0984, 0.1270, 0.4813, 0.1218, 0.1581, 0.1714, 0.1081], + device='cuda:1'), in_proj_covar=tensor([0.0530, 0.0440, 0.0433, 0.0542, 0.0428, 0.0449, 0.0429, 0.0389], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:35:16,518 INFO [train.py:901] (1/4) Epoch 22, batch 7450, loss[loss=0.1762, simple_loss=0.2582, pruned_loss=0.04705, over 7810.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2877, pruned_loss=0.0622, over 1608146.86 frames. ], batch size: 20, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:35:17,193 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.327e+02 2.972e+02 3.761e+02 7.589e+02, threshold=5.944e+02, percent-clipped=3.0 +2023-02-07 04:35:21,631 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177199.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:35:27,661 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 04:35:32,256 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177215.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:35:38,457 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177224.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:35:51,509 INFO [train.py:901] (1/4) Epoch 22, batch 7500, loss[loss=0.2006, simple_loss=0.2964, pruned_loss=0.05243, over 8779.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.288, pruned_loss=0.06208, over 1613385.34 frames. ], batch size: 50, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:35:58,950 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8667, 1.7164, 1.8780, 1.8024, 0.9798, 1.5662, 2.2093, 2.2216], + device='cuda:1'), covar=tensor([0.0453, 0.1222, 0.1674, 0.1326, 0.0626, 0.1500, 0.0627, 0.0565], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0159, 0.0099, 0.0162, 0.0111, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 04:36:23,738 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.80 vs. limit=5.0 +2023-02-07 04:36:25,359 INFO [train.py:901] (1/4) Epoch 22, batch 7550, loss[loss=0.1804, simple_loss=0.2793, pruned_loss=0.04076, over 8475.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2873, pruned_loss=0.06155, over 1611350.36 frames. ], batch size: 25, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:36:26,049 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.508e+02 3.019e+02 3.781e+02 7.904e+02, threshold=6.039e+02, percent-clipped=4.0 +2023-02-07 04:36:43,676 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4128, 1.2496, 2.1328, 1.1742, 1.9976, 2.2940, 2.4000, 1.9606], + device='cuda:1'), covar=tensor([0.0972, 0.1342, 0.0487, 0.1900, 0.0904, 0.0340, 0.0648, 0.0583], + device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0321, 0.0285, 0.0314, 0.0309, 0.0266, 0.0420, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 04:36:51,646 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177330.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:36:59,624 INFO [train.py:901] (1/4) Epoch 22, batch 7600, loss[loss=0.1595, simple_loss=0.2426, pruned_loss=0.03822, over 7921.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.288, pruned_loss=0.06173, over 1615051.94 frames. ], batch size: 20, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:37:11,460 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177358.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:37:29,851 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177383.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:37:35,779 INFO [train.py:901] (1/4) Epoch 22, batch 7650, loss[loss=0.2059, simple_loss=0.2877, pruned_loss=0.06206, over 8498.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2872, pruned_loss=0.06141, over 1615188.02 frames. ], batch size: 28, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:37:36,440 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.559e+02 3.074e+02 4.315e+02 1.263e+03, threshold=6.148e+02, percent-clipped=10.0 +2023-02-07 04:37:39,975 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177398.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:37:57,390 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177423.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:38:00,823 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6679, 2.6917, 1.9387, 2.3711, 2.4001, 1.6648, 2.2345, 2.2903], + device='cuda:1'), covar=tensor([0.1603, 0.0408, 0.1199, 0.0730, 0.0760, 0.1565, 0.1094, 0.1132], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0232, 0.0335, 0.0310, 0.0299, 0.0341, 0.0345, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 04:38:09,936 INFO [train.py:901] (1/4) Epoch 22, batch 7700, loss[loss=0.1884, simple_loss=0.2601, pruned_loss=0.05836, over 7708.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2863, pruned_loss=0.06109, over 1612231.08 frames. ], batch size: 18, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:38:30,444 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177471.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:38:31,729 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177473.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:38:38,598 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 04:38:41,834 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-02-07 04:38:45,982 INFO [train.py:901] (1/4) Epoch 22, batch 7750, loss[loss=0.2124, simple_loss=0.3039, pruned_loss=0.06049, over 8257.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2858, pruned_loss=0.06074, over 1609360.91 frames. ], batch size: 24, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:38:46,657 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.486e+02 3.125e+02 4.090e+02 1.041e+03, threshold=6.251e+02, percent-clipped=8.0 +2023-02-07 04:39:10,513 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6198, 1.8575, 1.9909, 1.1982, 2.1004, 1.3536, 0.6389, 1.7583], + device='cuda:1'), covar=tensor([0.0598, 0.0387, 0.0319, 0.0708, 0.0370, 0.1072, 0.0901, 0.0334], + device='cuda:1'), in_proj_covar=tensor([0.0452, 0.0392, 0.0344, 0.0446, 0.0378, 0.0534, 0.0390, 0.0422], + device='cuda:1'), out_proj_covar=tensor([1.2108e-04, 1.0291e-04, 9.0359e-05, 1.1744e-04, 9.9402e-05, 1.5060e-04, + 1.0499e-04, 1.1163e-04], device='cuda:1') +2023-02-07 04:39:20,411 INFO [train.py:901] (1/4) Epoch 22, batch 7800, loss[loss=0.1671, simple_loss=0.2508, pruned_loss=0.04171, over 7777.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2866, pruned_loss=0.06116, over 1615154.50 frames. ], batch size: 19, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:39:39,773 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177571.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:39:49,858 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:39:49,892 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:39:50,725 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-02-07 04:39:51,145 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177588.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:39:53,646 INFO [train.py:901] (1/4) Epoch 22, batch 7850, loss[loss=0.1765, simple_loss=0.269, pruned_loss=0.04195, over 8031.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2859, pruned_loss=0.06066, over 1614607.84 frames. ], batch size: 22, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:39:54,299 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.387e+02 2.753e+02 3.373e+02 6.542e+02, threshold=5.505e+02, percent-clipped=2.0 +2023-02-07 04:40:06,555 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177611.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:40:26,677 INFO [train.py:901] (1/4) Epoch 22, batch 7900, loss[loss=0.2267, simple_loss=0.3136, pruned_loss=0.06993, over 8291.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2858, pruned_loss=0.06074, over 1613458.40 frames. ], batch size: 23, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:40:53,550 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177682.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:41:00,029 INFO [train.py:901] (1/4) Epoch 22, batch 7950, loss[loss=0.1824, simple_loss=0.2665, pruned_loss=0.04911, over 7916.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2866, pruned_loss=0.06085, over 1616675.64 frames. ], batch size: 20, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:41:00,683 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.424e+02 2.966e+02 3.766e+02 9.319e+02, threshold=5.931e+02, percent-clipped=7.0 +2023-02-07 04:41:06,666 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7098, 5.7128, 5.1252, 2.7887, 5.1709, 5.4708, 5.2796, 5.2051], + device='cuda:1'), covar=tensor([0.0457, 0.0408, 0.0774, 0.3806, 0.0675, 0.0714, 0.1031, 0.0686], + device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0438, 0.0433, 0.0538, 0.0423, 0.0445, 0.0424, 0.0388], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:41:22,702 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8320, 1.4351, 1.6520, 1.3025, 0.8279, 1.4501, 1.5206, 1.5930], + device='cuda:1'), covar=tensor([0.0516, 0.1277, 0.1673, 0.1454, 0.0634, 0.1515, 0.0707, 0.0646], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0159, 0.0099, 0.0163, 0.0112, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 04:41:33,730 INFO [train.py:901] (1/4) Epoch 22, batch 8000, loss[loss=0.212, simple_loss=0.3067, pruned_loss=0.05866, over 8297.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2855, pruned_loss=0.06, over 1614682.40 frames. ], batch size: 23, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:42:06,690 INFO [train.py:901] (1/4) Epoch 22, batch 8050, loss[loss=0.1673, simple_loss=0.2503, pruned_loss=0.04211, over 7254.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2848, pruned_loss=0.06047, over 1600587.20 frames. ], batch size: 16, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:42:07,275 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.308e+02 2.923e+02 3.618e+02 1.070e+03, threshold=5.846e+02, percent-clipped=4.0 +2023-02-07 04:42:10,789 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177798.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:42:22,383 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177815.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:42:39,874 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 04:42:44,815 INFO [train.py:901] (1/4) Epoch 23, batch 0, loss[loss=0.2238, simple_loss=0.3032, pruned_loss=0.07221, over 8598.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3032, pruned_loss=0.07221, over 8598.00 frames. ], batch size: 31, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:42:44,816 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 04:42:56,157 INFO [train.py:935] (1/4) Epoch 23, validation: loss=0.1743, simple_loss=0.274, pruned_loss=0.0373, over 944034.00 frames. +2023-02-07 04:42:56,159 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 04:43:08,341 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177842.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:43:10,543 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177844.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:43:12,395 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 04:43:26,740 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177867.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:43:28,092 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177869.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:43:32,004 INFO [train.py:901] (1/4) Epoch 23, batch 50, loss[loss=0.2369, simple_loss=0.3145, pruned_loss=0.07965, over 8595.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2922, pruned_loss=0.06472, over 370800.27 frames. ], batch size: 39, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:43:42,573 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0451, 2.3649, 3.7032, 1.7839, 3.0542, 2.3080, 2.1510, 2.7324], + device='cuda:1'), covar=tensor([0.1600, 0.2106, 0.0834, 0.3812, 0.1499, 0.2890, 0.1834, 0.2342], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0605, 0.0556, 0.0648, 0.0648, 0.0594, 0.0538, 0.0633], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:43:45,276 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.650e+02 3.149e+02 3.939e+02 1.519e+03, threshold=6.298e+02, percent-clipped=14.0 +2023-02-07 04:43:46,682 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 04:44:01,101 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177915.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:44:02,080 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 04:44:07,959 INFO [train.py:901] (1/4) Epoch 23, batch 100, loss[loss=0.1799, simple_loss=0.2606, pruned_loss=0.04955, over 7793.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2901, pruned_loss=0.06266, over 646135.15 frames. ], batch size: 19, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:44:09,367 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 04:44:15,321 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0032, 2.3470, 1.9178, 2.9651, 1.4451, 1.6501, 2.1890, 2.2929], + device='cuda:1'), covar=tensor([0.0766, 0.0734, 0.0904, 0.0357, 0.1178, 0.1371, 0.0855, 0.0803], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0196, 0.0244, 0.0214, 0.0207, 0.0246, 0.0250, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 04:44:42,233 INFO [train.py:901] (1/4) Epoch 23, batch 150, loss[loss=0.1831, simple_loss=0.2789, pruned_loss=0.04364, over 8295.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2886, pruned_loss=0.06187, over 862789.72 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:44:49,532 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1896, 1.4341, 4.4015, 1.8497, 2.2191, 4.9235, 5.0549, 4.3051], + device='cuda:1'), covar=tensor([0.1246, 0.2138, 0.0267, 0.2219, 0.1532, 0.0176, 0.0351, 0.0556], + device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0323, 0.0287, 0.0318, 0.0313, 0.0269, 0.0424, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 04:44:54,933 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.352e+02 3.015e+02 3.767e+02 5.945e+02, threshold=6.031e+02, percent-clipped=0.0 +2023-02-07 04:44:55,345 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 04:45:18,306 INFO [train.py:901] (1/4) Epoch 23, batch 200, loss[loss=0.1879, simple_loss=0.2673, pruned_loss=0.05425, over 8238.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2889, pruned_loss=0.06253, over 1029136.88 frames. ], batch size: 22, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:45:19,114 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178026.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:45:21,851 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178030.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:45:53,024 INFO [train.py:901] (1/4) Epoch 23, batch 250, loss[loss=0.2734, simple_loss=0.3434, pruned_loss=0.1018, over 8200.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2875, pruned_loss=0.06179, over 1159292.44 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:46:04,759 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 04:46:06,099 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.380e+02 2.804e+02 3.484e+02 6.736e+02, threshold=5.609e+02, percent-clipped=2.0 +2023-02-07 04:46:12,814 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 04:46:28,460 INFO [train.py:901] (1/4) Epoch 23, batch 300, loss[loss=0.1731, simple_loss=0.2647, pruned_loss=0.04077, over 8106.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2877, pruned_loss=0.06209, over 1261928.42 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:46:40,068 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178141.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:46:40,620 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178142.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:46:45,435 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1224, 3.6074, 2.3452, 2.6207, 2.6870, 1.9069, 2.6600, 2.9092], + device='cuda:1'), covar=tensor([0.1492, 0.0366, 0.1069, 0.0826, 0.0810, 0.1498, 0.1036, 0.0980], + device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0231, 0.0332, 0.0308, 0.0297, 0.0337, 0.0342, 0.0313], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 04:46:52,875 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178159.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:46:53,826 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 04:46:54,583 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 04:46:59,072 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1362, 1.6863, 4.1329, 1.7896, 2.3931, 4.7180, 4.8299, 4.1093], + device='cuda:1'), covar=tensor([0.1307, 0.1890, 0.0303, 0.2150, 0.1335, 0.0196, 0.0397, 0.0546], + device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0322, 0.0288, 0.0316, 0.0312, 0.0267, 0.0423, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 04:47:03,747 INFO [train.py:901] (1/4) Epoch 23, batch 350, loss[loss=0.2214, simple_loss=0.3017, pruned_loss=0.07051, over 8604.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2879, pruned_loss=0.06184, over 1340958.62 frames. ], batch size: 31, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:47:16,039 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.418e+02 2.905e+02 3.451e+02 8.072e+02, threshold=5.809e+02, percent-clipped=5.0 +2023-02-07 04:47:23,332 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2700, 1.8703, 4.4463, 2.1711, 2.5901, 5.1145, 5.1284, 4.4719], + device='cuda:1'), covar=tensor([0.1185, 0.1641, 0.0253, 0.1750, 0.1100, 0.0173, 0.0428, 0.0522], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0320, 0.0286, 0.0315, 0.0311, 0.0266, 0.0421, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 04:47:38,679 INFO [train.py:901] (1/4) Epoch 23, batch 400, loss[loss=0.2095, simple_loss=0.3003, pruned_loss=0.0593, over 8034.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2875, pruned_loss=0.06191, over 1399593.31 frames. ], batch size: 22, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:47:43,337 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-07 04:47:51,120 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5305, 1.8296, 2.6912, 1.4601, 1.9569, 1.8570, 1.6801, 1.9567], + device='cuda:1'), covar=tensor([0.1959, 0.2669, 0.0987, 0.4606, 0.1887, 0.3385, 0.2304, 0.2286], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0605, 0.0556, 0.0645, 0.0647, 0.0592, 0.0537, 0.0633], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:48:02,280 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178257.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:48:15,034 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178274.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:48:15,486 INFO [train.py:901] (1/4) Epoch 23, batch 450, loss[loss=0.2072, simple_loss=0.2952, pruned_loss=0.05959, over 8087.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2876, pruned_loss=0.0617, over 1448100.35 frames. ], batch size: 21, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:48:15,622 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8788, 6.0130, 5.2800, 2.8368, 5.3416, 5.6900, 5.5260, 5.4228], + device='cuda:1'), covar=tensor([0.0560, 0.0355, 0.0810, 0.3937, 0.0687, 0.0879, 0.1007, 0.0552], + device='cuda:1'), in_proj_covar=tensor([0.0530, 0.0439, 0.0434, 0.0542, 0.0426, 0.0446, 0.0427, 0.0390], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:48:17,119 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0767, 2.4110, 1.9460, 2.9499, 1.4099, 1.7356, 2.1944, 2.3104], + device='cuda:1'), covar=tensor([0.0716, 0.0623, 0.0832, 0.0323, 0.1033, 0.1200, 0.0766, 0.0686], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0197, 0.0245, 0.0215, 0.0207, 0.0246, 0.0250, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 04:48:23,156 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178286.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:48:27,622 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.308e+02 2.812e+02 3.532e+02 1.107e+03, threshold=5.624e+02, percent-clipped=2.0 +2023-02-07 04:48:40,238 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178311.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:48:50,181 INFO [train.py:901] (1/4) Epoch 23, batch 500, loss[loss=0.2306, simple_loss=0.3064, pruned_loss=0.07739, over 8726.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2871, pruned_loss=0.0615, over 1483487.48 frames. ], batch size: 30, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:49:25,960 INFO [train.py:901] (1/4) Epoch 23, batch 550, loss[loss=0.1968, simple_loss=0.2788, pruned_loss=0.05741, over 8330.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2875, pruned_loss=0.06227, over 1508719.78 frames. ], batch size: 25, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:49:39,372 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.448e+02 3.105e+02 3.761e+02 9.562e+02, threshold=6.211e+02, percent-clipped=5.0 +2023-02-07 04:49:42,440 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178397.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:49:59,303 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178422.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:50:01,203 INFO [train.py:901] (1/4) Epoch 23, batch 600, loss[loss=0.2433, simple_loss=0.3294, pruned_loss=0.0786, over 8329.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2881, pruned_loss=0.06161, over 1536610.59 frames. ], batch size: 25, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:50:14,817 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 04:50:33,525 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178470.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:50:36,791 INFO [train.py:901] (1/4) Epoch 23, batch 650, loss[loss=0.2552, simple_loss=0.3367, pruned_loss=0.08691, over 8489.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2881, pruned_loss=0.06206, over 1552348.11 frames. ], batch size: 29, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:50:49,793 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.230e+02 2.701e+02 3.368e+02 8.641e+02, threshold=5.402e+02, percent-clipped=2.0 +2023-02-07 04:51:04,347 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178513.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:51:12,432 INFO [train.py:901] (1/4) Epoch 23, batch 700, loss[loss=0.1746, simple_loss=0.257, pruned_loss=0.04604, over 7664.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2891, pruned_loss=0.06272, over 1564279.32 frames. ], batch size: 19, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:51:16,067 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178530.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:51:18,198 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0631, 1.7874, 2.3033, 1.9844, 2.2421, 2.0988, 1.9034, 1.1783], + device='cuda:1'), covar=tensor([0.5909, 0.4871, 0.2078, 0.3828, 0.2582, 0.3189, 0.2035, 0.5339], + device='cuda:1'), in_proj_covar=tensor([0.0941, 0.0986, 0.0813, 0.0952, 0.0997, 0.0899, 0.0754, 0.0831], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 04:51:21,533 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178538.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:51:33,980 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178555.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:51:47,514 INFO [train.py:901] (1/4) Epoch 23, batch 750, loss[loss=0.1982, simple_loss=0.2681, pruned_loss=0.06414, over 7675.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2881, pruned_loss=0.06213, over 1575280.86 frames. ], batch size: 18, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:51:49,820 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7409, 1.6321, 2.4588, 1.5446, 1.3045, 2.3888, 0.5161, 1.5001], + device='cuda:1'), covar=tensor([0.1849, 0.1424, 0.0322, 0.1326, 0.2648, 0.0449, 0.2083, 0.1432], + device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0200, 0.0130, 0.0221, 0.0270, 0.0137, 0.0171, 0.0196], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 04:51:59,483 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5935, 1.8338, 2.6983, 1.4566, 1.9469, 1.9528, 1.6407, 2.0119], + device='cuda:1'), covar=tensor([0.1858, 0.2477, 0.0821, 0.4332, 0.1806, 0.3092, 0.2281, 0.2084], + device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0606, 0.0555, 0.0645, 0.0649, 0.0594, 0.0536, 0.0634], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:52:00,642 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.529e+02 2.988e+02 3.531e+02 9.866e+02, threshold=5.976e+02, percent-clipped=5.0 +2023-02-07 04:52:03,329 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 04:52:12,892 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 04:52:24,029 INFO [train.py:901] (1/4) Epoch 23, batch 800, loss[loss=0.187, simple_loss=0.2779, pruned_loss=0.04804, over 8287.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2886, pruned_loss=0.06212, over 1588827.14 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:52:32,115 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178637.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:52:38,309 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2398, 2.2630, 2.0524, 2.6398, 1.8978, 1.9926, 2.1832, 2.3677], + device='cuda:1'), covar=tensor([0.0582, 0.0656, 0.0713, 0.0456, 0.0828, 0.0952, 0.0653, 0.0602], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0197, 0.0245, 0.0214, 0.0206, 0.0245, 0.0249, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 04:52:57,754 INFO [train.py:901] (1/4) Epoch 23, batch 850, loss[loss=0.191, simple_loss=0.2926, pruned_loss=0.04476, over 8335.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2881, pruned_loss=0.06227, over 1594393.55 frames. ], batch size: 25, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:53:10,567 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.561e+02 2.992e+02 3.918e+02 1.040e+03, threshold=5.984e+02, percent-clipped=6.0 +2023-02-07 04:53:24,466 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178712.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:53:26,475 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178715.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 04:53:31,368 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2295, 3.1366, 2.9200, 1.4521, 2.8193, 2.9492, 2.8088, 2.8833], + device='cuda:1'), covar=tensor([0.1065, 0.0756, 0.1153, 0.5010, 0.1074, 0.1133, 0.1571, 0.0913], + device='cuda:1'), in_proj_covar=tensor([0.0535, 0.0440, 0.0435, 0.0544, 0.0426, 0.0448, 0.0431, 0.0389], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:53:34,044 INFO [train.py:901] (1/4) Epoch 23, batch 900, loss[loss=0.2343, simple_loss=0.3145, pruned_loss=0.07704, over 8451.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2874, pruned_loss=0.06189, over 1596261.64 frames. ], batch size: 50, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:53:55,223 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.83 vs. limit=5.0 +2023-02-07 04:54:03,494 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8678, 1.4861, 3.1683, 1.3946, 2.2564, 3.4260, 3.5830, 2.9186], + device='cuda:1'), covar=tensor([0.1184, 0.1784, 0.0358, 0.2195, 0.0987, 0.0250, 0.0533, 0.0560], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0321, 0.0286, 0.0315, 0.0310, 0.0267, 0.0422, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 04:54:09,420 INFO [train.py:901] (1/4) Epoch 23, batch 950, loss[loss=0.2123, simple_loss=0.3048, pruned_loss=0.05994, over 8106.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2881, pruned_loss=0.06187, over 1606764.36 frames. ], batch size: 23, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:54:18,540 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178788.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:54:21,847 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.663e+02 2.330e+02 2.907e+02 3.544e+02 9.473e+02, threshold=5.814e+02, percent-clipped=4.0 +2023-02-07 04:54:22,772 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0217, 1.2254, 1.2048, 0.5546, 1.2342, 1.0227, 0.0758, 1.1495], + device='cuda:1'), covar=tensor([0.0492, 0.0428, 0.0389, 0.0672, 0.0432, 0.1117, 0.0866, 0.0368], + device='cuda:1'), in_proj_covar=tensor([0.0453, 0.0395, 0.0347, 0.0448, 0.0380, 0.0536, 0.0393, 0.0425], + device='cuda:1'), out_proj_covar=tensor([1.2096e-04, 1.0363e-04, 9.1189e-05, 1.1773e-04, 9.9773e-05, 1.5107e-04, + 1.0597e-04, 1.1251e-04], device='cuda:1') +2023-02-07 04:54:35,788 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 04:54:37,113 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178814.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:54:43,576 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1823, 1.5775, 1.8350, 1.4387, 0.9969, 1.5373, 1.9662, 1.7773], + device='cuda:1'), covar=tensor([0.0517, 0.1278, 0.1640, 0.1447, 0.0621, 0.1485, 0.0632, 0.0635], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0159, 0.0100, 0.0163, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 04:54:45,361 INFO [train.py:901] (1/4) Epoch 23, batch 1000, loss[loss=0.228, simple_loss=0.3038, pruned_loss=0.07605, over 8500.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2878, pruned_loss=0.0619, over 1610919.94 frames. ], batch size: 26, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:54:49,035 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9317, 1.8040, 2.0337, 1.7557, 1.0100, 1.6567, 2.3437, 2.2029], + device='cuda:1'), covar=tensor([0.0448, 0.1156, 0.1595, 0.1350, 0.0581, 0.1393, 0.0562, 0.0575], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0159, 0.0100, 0.0163, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 04:55:12,368 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 04:55:21,365 INFO [train.py:901] (1/4) Epoch 23, batch 1050, loss[loss=0.1759, simple_loss=0.266, pruned_loss=0.04292, over 8483.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2875, pruned_loss=0.06149, over 1614048.28 frames. ], batch size: 25, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:55:25,394 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 04:55:33,395 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.332e+02 2.695e+02 3.454e+02 6.847e+02, threshold=5.390e+02, percent-clipped=5.0 +2023-02-07 04:55:46,652 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178912.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 04:55:50,005 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.5292, 5.5799, 4.9890, 2.6199, 4.9901, 5.2538, 5.1941, 5.0767], + device='cuda:1'), covar=tensor([0.0574, 0.0363, 0.0874, 0.4271, 0.0726, 0.0779, 0.0999, 0.0622], + device='cuda:1'), in_proj_covar=tensor([0.0532, 0.0440, 0.0435, 0.0543, 0.0427, 0.0448, 0.0431, 0.0388], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:55:56,201 INFO [train.py:901] (1/4) Epoch 23, batch 1100, loss[loss=0.2226, simple_loss=0.2953, pruned_loss=0.07497, over 8506.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2881, pruned_loss=0.0617, over 1615756.59 frames. ], batch size: 26, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:55:59,138 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178929.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:56:01,239 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2311, 1.2492, 3.3665, 1.0344, 2.9753, 2.8161, 3.0878, 2.9844], + device='cuda:1'), covar=tensor([0.0869, 0.4485, 0.0857, 0.4433, 0.1512, 0.1229, 0.0822, 0.1008], + device='cuda:1'), in_proj_covar=tensor([0.0634, 0.0644, 0.0700, 0.0630, 0.0711, 0.0604, 0.0603, 0.0683], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:56:32,124 INFO [train.py:901] (1/4) Epoch 23, batch 1150, loss[loss=0.184, simple_loss=0.268, pruned_loss=0.05004, over 7938.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2883, pruned_loss=0.06194, over 1617015.51 frames. ], batch size: 20, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:56:36,260 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 04:56:36,327 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178981.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:56:45,240 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.628e+02 3.162e+02 4.177e+02 1.087e+03, threshold=6.324e+02, percent-clipped=6.0 +2023-02-07 04:57:07,132 INFO [train.py:901] (1/4) Epoch 23, batch 1200, loss[loss=0.2559, simple_loss=0.322, pruned_loss=0.09491, over 7046.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2886, pruned_loss=0.06216, over 1613128.58 frames. ], batch size: 72, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:57:13,472 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3603, 2.1229, 2.7998, 2.2333, 2.6609, 2.3656, 2.1850, 1.4969], + device='cuda:1'), covar=tensor([0.5257, 0.4782, 0.1994, 0.3543, 0.2509, 0.2926, 0.1832, 0.5095], + device='cuda:1'), in_proj_covar=tensor([0.0935, 0.0981, 0.0807, 0.0947, 0.0992, 0.0897, 0.0751, 0.0826], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 04:57:19,757 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6175, 1.3863, 1.6701, 1.2940, 0.8603, 1.4834, 1.4738, 1.5378], + device='cuda:1'), covar=tensor([0.0583, 0.1290, 0.1640, 0.1462, 0.0598, 0.1475, 0.0693, 0.0595], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0159, 0.0099, 0.0162, 0.0111, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 04:57:29,070 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179056.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:57:31,018 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179059.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 04:57:42,798 INFO [train.py:901] (1/4) Epoch 23, batch 1250, loss[loss=0.1919, simple_loss=0.2776, pruned_loss=0.05311, over 8327.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2881, pruned_loss=0.06162, over 1615556.74 frames. ], batch size: 25, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:57:55,995 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 2.289e+02 2.896e+02 3.686e+02 5.954e+02, threshold=5.791e+02, percent-clipped=0.0 +2023-02-07 04:57:58,269 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179096.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:58:19,004 INFO [train.py:901] (1/4) Epoch 23, batch 1300, loss[loss=0.1894, simple_loss=0.2786, pruned_loss=0.05009, over 8104.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2864, pruned_loss=0.06044, over 1616308.76 frames. ], batch size: 23, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:58:24,098 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179132.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:58:51,891 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179171.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:58:53,994 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179174.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 04:58:54,483 INFO [train.py:901] (1/4) Epoch 23, batch 1350, loss[loss=0.1795, simple_loss=0.262, pruned_loss=0.04851, over 7528.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2867, pruned_loss=0.06011, over 1619433.24 frames. ], batch size: 18, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:59:01,704 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179185.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:59:07,800 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 2.184e+02 2.635e+02 3.098e+02 5.270e+02, threshold=5.271e+02, percent-clipped=0.0 +2023-02-07 04:59:20,397 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179210.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:59:30,634 INFO [train.py:901] (1/4) Epoch 23, batch 1400, loss[loss=0.2008, simple_loss=0.2942, pruned_loss=0.05373, over 8351.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2874, pruned_loss=0.06019, over 1618578.41 frames. ], batch size: 24, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:59:41,531 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 04:59:42,859 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6563, 2.2874, 4.1220, 1.5036, 3.0259, 2.3260, 1.7609, 2.8500], + device='cuda:1'), covar=tensor([0.1973, 0.2703, 0.0861, 0.4642, 0.1822, 0.3199, 0.2364, 0.2401], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0607, 0.0555, 0.0647, 0.0651, 0.0594, 0.0537, 0.0633], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:59:44,216 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9931, 1.4380, 1.7508, 1.3450, 1.0588, 1.5056, 1.7415, 1.7584], + device='cuda:1'), covar=tensor([0.0584, 0.1325, 0.1665, 0.1498, 0.0630, 0.1479, 0.0722, 0.0625], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0189, 0.0158, 0.0100, 0.0162, 0.0111, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 04:59:44,934 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5365, 1.6098, 4.7228, 1.7072, 4.1642, 3.9327, 4.2913, 4.1459], + device='cuda:1'), covar=tensor([0.0581, 0.4625, 0.0442, 0.3997, 0.1081, 0.0907, 0.0550, 0.0647], + device='cuda:1'), in_proj_covar=tensor([0.0630, 0.0638, 0.0697, 0.0629, 0.0707, 0.0604, 0.0601, 0.0677], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 04:59:47,170 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179247.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:59:52,290 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 +2023-02-07 04:59:53,422 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179256.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:00:06,546 INFO [train.py:901] (1/4) Epoch 23, batch 1450, loss[loss=0.2006, simple_loss=0.2785, pruned_loss=0.06138, over 7656.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2876, pruned_loss=0.06032, over 1619464.06 frames. ], batch size: 19, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:00:16,912 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 05:00:18,543 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6917, 1.4168, 1.6144, 1.3578, 1.0034, 1.4317, 1.6207, 1.6817], + device='cuda:1'), covar=tensor([0.0557, 0.1286, 0.1605, 0.1387, 0.0607, 0.1477, 0.0686, 0.0573], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0189, 0.0158, 0.0100, 0.0162, 0.0111, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 05:00:19,763 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.293e+02 2.971e+02 3.774e+02 8.745e+02, threshold=5.941e+02, percent-clipped=9.0 +2023-02-07 05:00:43,625 INFO [train.py:901] (1/4) Epoch 23, batch 1500, loss[loss=0.2516, simple_loss=0.3268, pruned_loss=0.0882, over 8371.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.289, pruned_loss=0.06115, over 1621427.01 frames. ], batch size: 49, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:01:03,401 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179352.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:01:16,378 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179371.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:01:18,843 INFO [train.py:901] (1/4) Epoch 23, batch 1550, loss[loss=0.1922, simple_loss=0.2884, pruned_loss=0.04798, over 8359.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2896, pruned_loss=0.06166, over 1622583.03 frames. ], batch size: 24, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:01:20,471 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179377.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:01:21,725 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179379.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:01:23,199 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3106, 1.4575, 1.3174, 1.7702, 0.6570, 1.1753, 1.2691, 1.4411], + device='cuda:1'), covar=tensor([0.0933, 0.0804, 0.0940, 0.0521, 0.1186, 0.1401, 0.0826, 0.0793], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0198, 0.0244, 0.0214, 0.0206, 0.0247, 0.0250, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 05:01:28,617 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7274, 5.8222, 5.1640, 2.5830, 5.2087, 5.4866, 5.3837, 5.3330], + device='cuda:1'), covar=tensor([0.0566, 0.0371, 0.0826, 0.4265, 0.0735, 0.0924, 0.1105, 0.0573], + device='cuda:1'), in_proj_covar=tensor([0.0531, 0.0439, 0.0434, 0.0543, 0.0428, 0.0449, 0.0431, 0.0388], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:01:31,097 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.349e+02 2.958e+02 3.969e+02 7.808e+02, threshold=5.916e+02, percent-clipped=1.0 +2023-02-07 05:01:54,009 INFO [train.py:901] (1/4) Epoch 23, batch 1600, loss[loss=0.2036, simple_loss=0.2898, pruned_loss=0.05867, over 8463.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2887, pruned_loss=0.0616, over 1623620.30 frames. ], batch size: 25, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:01:56,465 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179427.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:01:58,507 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179430.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:02:14,218 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-07 05:02:14,544 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179452.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:02:16,577 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179455.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:02:21,379 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179462.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:02:31,191 INFO [train.py:901] (1/4) Epoch 23, batch 1650, loss[loss=0.219, simple_loss=0.3075, pruned_loss=0.06519, over 8105.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2882, pruned_loss=0.0609, over 1625525.02 frames. ], batch size: 23, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:02:43,578 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.367e+02 2.783e+02 3.381e+02 8.055e+02, threshold=5.566e+02, percent-clipped=4.0 +2023-02-07 05:02:50,700 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179503.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:03:06,279 INFO [train.py:901] (1/4) Epoch 23, batch 1700, loss[loss=0.1986, simple_loss=0.2801, pruned_loss=0.05859, over 8087.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2882, pruned_loss=0.06113, over 1625850.14 frames. ], batch size: 21, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:03:08,709 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179528.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:03:42,324 INFO [train.py:901] (1/4) Epoch 23, batch 1750, loss[loss=0.1942, simple_loss=0.2902, pruned_loss=0.04907, over 8752.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2886, pruned_loss=0.06153, over 1620066.43 frames. ], batch size: 30, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 05:03:48,660 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7642, 1.6283, 2.2812, 1.5983, 1.2946, 2.2589, 0.8263, 1.5805], + device='cuda:1'), covar=tensor([0.1737, 0.1158, 0.0360, 0.1028, 0.2473, 0.0420, 0.1693, 0.1208], + device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0198, 0.0129, 0.0219, 0.0267, 0.0136, 0.0169, 0.0192], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 05:03:56,238 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.481e+02 2.857e+02 3.517e+02 8.396e+02, threshold=5.713e+02, percent-clipped=3.0 +2023-02-07 05:04:17,975 INFO [train.py:901] (1/4) Epoch 23, batch 1800, loss[loss=0.2588, simple_loss=0.33, pruned_loss=0.09381, over 8491.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2874, pruned_loss=0.06108, over 1619844.97 frames. ], batch size: 26, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 05:04:19,569 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179627.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:04:37,265 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179652.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:04:38,540 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179654.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:04:54,468 INFO [train.py:901] (1/4) Epoch 23, batch 1850, loss[loss=0.195, simple_loss=0.2827, pruned_loss=0.05359, over 8586.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2862, pruned_loss=0.06085, over 1617832.92 frames. ], batch size: 31, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 05:04:58,795 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5322, 1.9735, 2.9397, 1.3806, 2.2723, 1.8263, 1.7432, 2.1870], + device='cuda:1'), covar=tensor([0.1885, 0.2519, 0.1043, 0.4412, 0.1881, 0.3308, 0.2241, 0.2408], + device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0610, 0.0558, 0.0648, 0.0653, 0.0596, 0.0540, 0.0634], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:05:07,479 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.311e+02 2.831e+02 3.615e+02 8.108e+02, threshold=5.663e+02, percent-clipped=6.0 +2023-02-07 05:05:12,872 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 05:05:28,483 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:05:29,798 INFO [train.py:901] (1/4) Epoch 23, batch 1900, loss[loss=0.1898, simple_loss=0.2715, pruned_loss=0.05408, over 8245.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.287, pruned_loss=0.06145, over 1618524.91 frames. ], batch size: 22, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 05:05:41,138 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-07 05:05:59,886 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 05:06:05,579 INFO [train.py:901] (1/4) Epoch 23, batch 1950, loss[loss=0.2272, simple_loss=0.3087, pruned_loss=0.07282, over 8294.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.287, pruned_loss=0.06185, over 1617328.20 frames. ], batch size: 23, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:06:12,661 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 05:06:19,490 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.457e+02 2.986e+02 3.643e+02 8.972e+02, threshold=5.972e+02, percent-clipped=4.0 +2023-02-07 05:06:22,396 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7371, 4.7654, 4.2850, 2.1143, 4.2275, 4.3469, 4.3019, 4.2614], + device='cuda:1'), covar=tensor([0.0651, 0.0474, 0.1028, 0.4331, 0.0847, 0.0743, 0.1139, 0.0601], + device='cuda:1'), in_proj_covar=tensor([0.0534, 0.0441, 0.0437, 0.0545, 0.0431, 0.0453, 0.0434, 0.0391], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:06:28,061 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179806.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:06:31,279 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 05:06:34,176 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179814.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:06:41,618 INFO [train.py:901] (1/4) Epoch 23, batch 2000, loss[loss=0.191, simple_loss=0.2728, pruned_loss=0.0546, over 8334.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2861, pruned_loss=0.06138, over 1615059.58 frames. ], batch size: 25, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:06:43,151 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1720, 1.4653, 1.8351, 1.4877, 0.9301, 1.5538, 1.8383, 1.6951], + device='cuda:1'), covar=tensor([0.0484, 0.1274, 0.1584, 0.1365, 0.0579, 0.1380, 0.0636, 0.0612], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0189, 0.0159, 0.0100, 0.0162, 0.0112, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 05:06:50,599 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179838.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:07:15,518 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-07 05:07:16,430 INFO [train.py:901] (1/4) Epoch 23, batch 2050, loss[loss=0.246, simple_loss=0.3202, pruned_loss=0.08592, over 6823.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2869, pruned_loss=0.06185, over 1614580.37 frames. ], batch size: 71, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:07:30,035 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.444e+02 2.856e+02 3.794e+02 1.051e+03, threshold=5.713e+02, percent-clipped=7.0 +2023-02-07 05:07:40,337 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-07 05:07:49,563 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179921.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:07:51,623 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179924.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:07:52,156 INFO [train.py:901] (1/4) Epoch 23, batch 2100, loss[loss=0.1857, simple_loss=0.2768, pruned_loss=0.04737, over 7649.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2876, pruned_loss=0.06202, over 1615998.33 frames. ], batch size: 19, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:07:55,136 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7494, 1.8430, 1.6700, 2.3270, 0.9734, 1.4376, 1.7764, 1.9331], + device='cuda:1'), covar=tensor([0.0791, 0.0783, 0.0883, 0.0436, 0.1118, 0.1378, 0.0731, 0.0683], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0197, 0.0243, 0.0213, 0.0206, 0.0246, 0.0249, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 05:08:00,347 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.85 vs. limit=5.0 +2023-02-07 05:08:04,834 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179942.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:08:05,925 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.31 vs. limit=5.0 +2023-02-07 05:08:27,555 INFO [train.py:901] (1/4) Epoch 23, batch 2150, loss[loss=0.2163, simple_loss=0.2966, pruned_loss=0.06794, over 7818.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2879, pruned_loss=0.06193, over 1615581.75 frames. ], batch size: 20, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:08:41,574 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.311e+02 2.940e+02 3.642e+02 8.826e+02, threshold=5.880e+02, percent-clipped=6.0 +2023-02-07 05:08:44,583 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179998.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:09:05,721 INFO [train.py:901] (1/4) Epoch 23, batch 2200, loss[loss=0.216, simple_loss=0.3016, pruned_loss=0.06517, over 8021.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2867, pruned_loss=0.0613, over 1612993.74 frames. ], batch size: 22, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:09:13,519 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2733, 1.6104, 1.2836, 2.6436, 1.1043, 1.1860, 1.9103, 1.8802], + device='cuda:1'), covar=tensor([0.1536, 0.1310, 0.1945, 0.0383, 0.1406, 0.2011, 0.0935, 0.0977], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0196, 0.0242, 0.0213, 0.0205, 0.0245, 0.0249, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 05:09:18,800 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180044.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:09:40,651 INFO [train.py:901] (1/4) Epoch 23, batch 2250, loss[loss=0.2327, simple_loss=0.3049, pruned_loss=0.08026, over 8199.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2867, pruned_loss=0.06092, over 1615705.87 frames. ], batch size: 23, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:09:53,773 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.578e+02 2.377e+02 2.815e+02 3.570e+02 6.536e+02, threshold=5.630e+02, percent-clipped=1.0 +2023-02-07 05:09:54,034 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180094.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:10:07,866 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180113.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:10:12,125 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180119.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:10:16,776 INFO [train.py:901] (1/4) Epoch 23, batch 2300, loss[loss=0.2525, simple_loss=0.3186, pruned_loss=0.09326, over 7189.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.287, pruned_loss=0.06097, over 1616462.64 frames. ], batch size: 71, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:10:40,109 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180158.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:10:41,670 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3739, 1.6363, 1.6747, 1.1425, 1.7046, 1.3733, 0.3316, 1.6390], + device='cuda:1'), covar=tensor([0.0476, 0.0357, 0.0282, 0.0481, 0.0433, 0.0910, 0.0824, 0.0258], + device='cuda:1'), in_proj_covar=tensor([0.0453, 0.0391, 0.0346, 0.0446, 0.0380, 0.0534, 0.0392, 0.0420], + device='cuda:1'), out_proj_covar=tensor([1.2101e-04, 1.0256e-04, 9.0863e-05, 1.1726e-04, 9.9980e-05, 1.5037e-04, + 1.0566e-04, 1.1127e-04], device='cuda:1') +2023-02-07 05:10:52,673 INFO [train.py:901] (1/4) Epoch 23, batch 2350, loss[loss=0.2214, simple_loss=0.317, pruned_loss=0.06296, over 8357.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2868, pruned_loss=0.06086, over 1617687.49 frames. ], batch size: 24, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:10:54,342 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180177.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:11:05,883 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.451e+02 2.928e+02 3.544e+02 9.883e+02, threshold=5.856e+02, percent-clipped=4.0 +2023-02-07 05:11:11,595 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180202.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:11:25,826 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180223.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:11:27,102 INFO [train.py:901] (1/4) Epoch 23, batch 2400, loss[loss=0.1836, simple_loss=0.2581, pruned_loss=0.05456, over 7421.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2875, pruned_loss=0.06149, over 1618642.33 frames. ], batch size: 17, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:11:57,640 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8750, 1.4823, 1.8258, 1.4662, 0.9139, 1.5791, 1.7543, 1.7281], + device='cuda:1'), covar=tensor([0.0503, 0.1215, 0.1570, 0.1380, 0.0583, 0.1370, 0.0650, 0.0583], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0159, 0.0100, 0.0162, 0.0111, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 05:11:59,472 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180268.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:12:02,929 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180273.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:12:04,119 INFO [train.py:901] (1/4) Epoch 23, batch 2450, loss[loss=0.2131, simple_loss=0.3023, pruned_loss=0.0619, over 8324.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2877, pruned_loss=0.06129, over 1621304.46 frames. ], batch size: 25, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:12:12,625 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180286.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:12:18,007 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.501e+02 2.918e+02 3.866e+02 1.157e+03, threshold=5.835e+02, percent-clipped=6.0 +2023-02-07 05:12:24,145 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 05:12:39,634 INFO [train.py:901] (1/4) Epoch 23, batch 2500, loss[loss=0.1916, simple_loss=0.2804, pruned_loss=0.05145, over 8203.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2878, pruned_loss=0.06112, over 1619761.26 frames. ], batch size: 23, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:13:00,496 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180354.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:12,920 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180369.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:16,778 INFO [train.py:901] (1/4) Epoch 23, batch 2550, loss[loss=0.2129, simple_loss=0.2822, pruned_loss=0.07186, over 7195.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2871, pruned_loss=0.06078, over 1619022.32 frames. ], batch size: 16, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:13:21,734 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0254, 2.1788, 1.8665, 2.7681, 1.2693, 1.6365, 1.9869, 2.2328], + device='cuda:1'), covar=tensor([0.0703, 0.0760, 0.0860, 0.0375, 0.1103, 0.1262, 0.0767, 0.0747], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0198, 0.0243, 0.0214, 0.0206, 0.0246, 0.0249, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 05:13:22,405 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180383.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:13:22,606 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-07 05:13:25,751 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180388.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:29,875 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.435e+02 3.031e+02 3.942e+02 1.076e+03, threshold=6.063e+02, percent-clipped=1.0 +2023-02-07 05:13:30,131 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180394.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:32,990 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2372, 2.1168, 2.7353, 2.1997, 2.6991, 2.3333, 2.1466, 1.4706], + device='cuda:1'), covar=tensor([0.5854, 0.5052, 0.2056, 0.4038, 0.2568, 0.3356, 0.2021, 0.5554], + device='cuda:1'), in_proj_covar=tensor([0.0948, 0.0993, 0.0818, 0.0954, 0.1003, 0.0905, 0.0756, 0.0835], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 05:13:35,752 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180401.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:35,901 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 05:13:48,409 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-07 05:13:51,963 INFO [train.py:901] (1/4) Epoch 23, batch 2600, loss[loss=0.1708, simple_loss=0.2534, pruned_loss=0.04413, over 7518.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2871, pruned_loss=0.06057, over 1613984.64 frames. ], batch size: 18, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:14:28,401 INFO [train.py:901] (1/4) Epoch 23, batch 2650, loss[loss=0.1756, simple_loss=0.2411, pruned_loss=0.05505, over 7228.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2863, pruned_loss=0.06019, over 1613327.61 frames. ], batch size: 16, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:14:42,183 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 2.331e+02 2.876e+02 3.734e+02 9.435e+02, threshold=5.753e+02, percent-clipped=4.0 +2023-02-07 05:14:48,440 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180503.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:04,100 INFO [train.py:901] (1/4) Epoch 23, batch 2700, loss[loss=0.2554, simple_loss=0.3276, pruned_loss=0.09157, over 8578.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2864, pruned_loss=0.0604, over 1614578.34 frames. ], batch size: 31, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:15:07,063 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180529.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:24,146 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180554.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:33,270 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180567.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:39,679 INFO [train.py:901] (1/4) Epoch 23, batch 2750, loss[loss=0.1809, simple_loss=0.2713, pruned_loss=0.04523, over 7816.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2873, pruned_loss=0.06067, over 1619754.51 frames. ], batch size: 20, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:15:48,808 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180588.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:53,488 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.355e+02 2.814e+02 3.432e+02 9.125e+02, threshold=5.629e+02, percent-clipped=4.0 +2023-02-07 05:16:15,665 INFO [train.py:901] (1/4) Epoch 23, batch 2800, loss[loss=0.2165, simple_loss=0.3056, pruned_loss=0.06366, over 8500.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2868, pruned_loss=0.06093, over 1617190.37 frames. ], batch size: 26, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:16:25,073 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-07 05:16:26,282 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180639.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:16:38,613 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180657.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:16:43,305 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180664.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:16:50,738 INFO [train.py:901] (1/4) Epoch 23, batch 2850, loss[loss=0.2074, simple_loss=0.2946, pruned_loss=0.0601, over 8030.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.287, pruned_loss=0.06124, over 1615349.24 frames. ], batch size: 22, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:16:55,705 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:16:55,727 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:17:04,507 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.469e+02 3.037e+02 3.866e+02 9.714e+02, threshold=6.075e+02, percent-clipped=7.0 +2023-02-07 05:17:07,438 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180698.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:17:27,378 INFO [train.py:901] (1/4) Epoch 23, batch 2900, loss[loss=0.2079, simple_loss=0.288, pruned_loss=0.06395, over 8446.00 frames. ], tot_loss[loss=0.205, simple_loss=0.287, pruned_loss=0.06145, over 1611855.43 frames. ], batch size: 29, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:17:45,110 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180750.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:17:52,130 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180759.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:17:56,171 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5770, 1.8861, 2.0181, 1.1617, 2.1247, 1.4448, 0.5044, 1.8516], + device='cuda:1'), covar=tensor([0.0613, 0.0346, 0.0258, 0.0651, 0.0449, 0.0890, 0.0935, 0.0311], + device='cuda:1'), in_proj_covar=tensor([0.0455, 0.0393, 0.0347, 0.0446, 0.0379, 0.0535, 0.0393, 0.0422], + device='cuda:1'), out_proj_covar=tensor([1.2156e-04, 1.0301e-04, 9.1094e-05, 1.1721e-04, 9.9850e-05, 1.5056e-04, + 1.0574e-04, 1.1164e-04], device='cuda:1') +2023-02-07 05:17:59,461 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 05:18:02,964 INFO [train.py:901] (1/4) Epoch 23, batch 2950, loss[loss=0.2145, simple_loss=0.2989, pruned_loss=0.06508, over 8239.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.287, pruned_loss=0.06175, over 1605014.16 frames. ], batch size: 22, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:18:09,314 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180784.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:18:16,020 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.356e+02 2.925e+02 3.942e+02 6.480e+02, threshold=5.850e+02, percent-clipped=1.0 +2023-02-07 05:18:24,006 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7189, 2.3390, 3.8379, 1.8215, 1.6333, 3.6547, 0.6321, 2.1528], + device='cuda:1'), covar=tensor([0.1819, 0.1286, 0.0199, 0.1926, 0.2862, 0.0342, 0.2239, 0.1546], + device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0199, 0.0129, 0.0222, 0.0271, 0.0137, 0.0171, 0.0195], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 05:18:30,325 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180813.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:18:34,562 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.87 vs. limit=5.0 +2023-02-07 05:18:38,135 INFO [train.py:901] (1/4) Epoch 23, batch 3000, loss[loss=0.1722, simple_loss=0.2716, pruned_loss=0.03638, over 8240.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2873, pruned_loss=0.0616, over 1607784.73 frames. ], batch size: 24, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:18:38,135 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 05:18:50,537 INFO [train.py:935] (1/4) Epoch 23, validation: loss=0.1735, simple_loss=0.2731, pruned_loss=0.03696, over 944034.00 frames. +2023-02-07 05:18:50,538 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 05:19:03,707 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:19:26,990 INFO [train.py:901] (1/4) Epoch 23, batch 3050, loss[loss=0.2226, simple_loss=0.3089, pruned_loss=0.06816, over 8340.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2872, pruned_loss=0.06149, over 1608423.73 frames. ], batch size: 26, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:19:33,846 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-07 05:19:40,678 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 2.515e+02 3.107e+02 3.968e+02 1.139e+03, threshold=6.214e+02, percent-clipped=7.0 +2023-02-07 05:20:02,332 INFO [train.py:901] (1/4) Epoch 23, batch 3100, loss[loss=0.2189, simple_loss=0.2961, pruned_loss=0.07092, over 8788.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2877, pruned_loss=0.06156, over 1610747.51 frames. ], batch size: 30, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:20:07,185 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180932.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:20:11,450 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180938.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:20:22,461 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9083, 5.9234, 5.2126, 2.8847, 5.3181, 5.6873, 5.4527, 5.4539], + device='cuda:1'), covar=tensor([0.0437, 0.0326, 0.0774, 0.3818, 0.0606, 0.0610, 0.0915, 0.0539], + device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0438, 0.0431, 0.0540, 0.0429, 0.0445, 0.0430, 0.0389], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:20:29,348 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180963.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:20:38,173 INFO [train.py:901] (1/4) Epoch 23, batch 3150, loss[loss=0.1874, simple_loss=0.288, pruned_loss=0.04342, over 8500.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2874, pruned_loss=0.06153, over 1608036.56 frames. ], batch size: 26, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:20:51,967 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 2.467e+02 3.042e+02 3.660e+02 1.036e+03, threshold=6.084e+02, percent-clipped=2.0 +2023-02-07 05:20:53,523 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5491, 1.9204, 2.0027, 1.2952, 2.1413, 1.5857, 0.5610, 1.8751], + device='cuda:1'), covar=tensor([0.0628, 0.0353, 0.0253, 0.0610, 0.0387, 0.0876, 0.0914, 0.0282], + device='cuda:1'), in_proj_covar=tensor([0.0455, 0.0392, 0.0345, 0.0446, 0.0378, 0.0533, 0.0391, 0.0420], + device='cuda:1'), out_proj_covar=tensor([1.2148e-04, 1.0270e-04, 9.0483e-05, 1.1729e-04, 9.9306e-05, 1.4992e-04, + 1.0529e-04, 1.1097e-04], device='cuda:1') +2023-02-07 05:21:14,460 INFO [train.py:901] (1/4) Epoch 23, batch 3200, loss[loss=0.2289, simple_loss=0.3087, pruned_loss=0.07462, over 8354.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2883, pruned_loss=0.06152, over 1614757.00 frames. ], batch size: 24, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:21:18,874 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3164, 1.5824, 1.6629, 1.1260, 1.6940, 1.3526, 0.3020, 1.6428], + device='cuda:1'), covar=tensor([0.0439, 0.0341, 0.0288, 0.0456, 0.0403, 0.0857, 0.0765, 0.0249], + device='cuda:1'), in_proj_covar=tensor([0.0453, 0.0392, 0.0345, 0.0446, 0.0378, 0.0532, 0.0391, 0.0419], + device='cuda:1'), out_proj_covar=tensor([1.2117e-04, 1.0258e-04, 9.0443e-05, 1.1715e-04, 9.9282e-05, 1.4990e-04, + 1.0525e-04, 1.1091e-04], device='cuda:1') +2023-02-07 05:21:29,942 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181047.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:21:45,859 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181069.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:21:49,893 INFO [train.py:901] (1/4) Epoch 23, batch 3250, loss[loss=0.1668, simple_loss=0.2614, pruned_loss=0.03608, over 8466.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2864, pruned_loss=0.06045, over 1612161.76 frames. ], batch size: 25, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:22:03,763 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.376e+02 2.917e+02 3.369e+02 6.745e+02, threshold=5.834e+02, percent-clipped=1.0 +2023-02-07 05:22:03,877 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=181094.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:22:04,011 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181094.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:22:26,227 INFO [train.py:901] (1/4) Epoch 23, batch 3300, loss[loss=0.181, simple_loss=0.2504, pruned_loss=0.05577, over 7708.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2871, pruned_loss=0.06079, over 1613627.31 frames. ], batch size: 18, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:23:01,366 INFO [train.py:901] (1/4) Epoch 23, batch 3350, loss[loss=0.2375, simple_loss=0.313, pruned_loss=0.08103, over 8617.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2873, pruned_loss=0.06103, over 1614568.35 frames. ], batch size: 49, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:23:10,453 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=181187.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:23:14,980 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.358e+02 3.053e+02 3.666e+02 9.674e+02, threshold=6.107e+02, percent-clipped=1.0 +2023-02-07 05:23:26,411 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181209.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:23:38,227 INFO [train.py:901] (1/4) Epoch 23, batch 3400, loss[loss=0.2131, simple_loss=0.3018, pruned_loss=0.06217, over 8108.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2878, pruned_loss=0.06171, over 1615371.55 frames. ], batch size: 23, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:23:55,608 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3642, 2.1088, 2.7779, 2.3249, 2.6058, 2.4109, 2.1935, 1.4884], + device='cuda:1'), covar=tensor([0.5328, 0.4759, 0.1746, 0.3373, 0.2396, 0.2779, 0.1785, 0.5086], + device='cuda:1'), in_proj_covar=tensor([0.0944, 0.0987, 0.0811, 0.0954, 0.0998, 0.0902, 0.0753, 0.0830], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 05:24:04,473 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7128, 1.9709, 2.1454, 1.4186, 2.2681, 1.5712, 0.6536, 1.9679], + device='cuda:1'), covar=tensor([0.0624, 0.0375, 0.0299, 0.0643, 0.0431, 0.0923, 0.0913, 0.0320], + device='cuda:1'), in_proj_covar=tensor([0.0453, 0.0392, 0.0345, 0.0448, 0.0379, 0.0534, 0.0391, 0.0421], + device='cuda:1'), out_proj_covar=tensor([1.2093e-04, 1.0258e-04, 9.0627e-05, 1.1771e-04, 9.9600e-05, 1.5055e-04, + 1.0540e-04, 1.1130e-04], device='cuda:1') +2023-02-07 05:24:13,226 INFO [train.py:901] (1/4) Epoch 23, batch 3450, loss[loss=0.2464, simple_loss=0.3193, pruned_loss=0.08675, over 8509.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2881, pruned_loss=0.06256, over 1612234.84 frames. ], batch size: 26, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:24:27,416 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.466e+02 2.960e+02 3.783e+02 8.296e+02, threshold=5.920e+02, percent-clipped=4.0 +2023-02-07 05:24:29,052 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8675, 1.7566, 2.7410, 2.2289, 2.4296, 1.9243, 1.6259, 1.2888], + device='cuda:1'), covar=tensor([0.6701, 0.6057, 0.1901, 0.3597, 0.2800, 0.3972, 0.2925, 0.5232], + device='cuda:1'), in_proj_covar=tensor([0.0943, 0.0988, 0.0811, 0.0954, 0.0998, 0.0901, 0.0754, 0.0830], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 05:24:32,993 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181302.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:24:33,730 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181303.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:24:42,795 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181315.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:24:49,550 INFO [train.py:901] (1/4) Epoch 23, batch 3500, loss[loss=0.1936, simple_loss=0.2803, pruned_loss=0.05345, over 8467.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2875, pruned_loss=0.06191, over 1616434.72 frames. ], batch size: 25, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:24:52,759 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181328.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:25:04,317 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([6.0298, 1.6438, 6.0965, 2.2129, 5.5235, 5.0960, 5.6356, 5.5251], + device='cuda:1'), covar=tensor([0.0348, 0.4771, 0.0338, 0.3791, 0.0863, 0.0798, 0.0407, 0.0450], + device='cuda:1'), in_proj_covar=tensor([0.0639, 0.0648, 0.0705, 0.0637, 0.0715, 0.0612, 0.0609, 0.0685], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:25:07,625 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 05:25:25,296 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2044, 4.0026, 2.5114, 3.0954, 3.0421, 2.2031, 3.0388, 3.1930], + device='cuda:1'), covar=tensor([0.1606, 0.0290, 0.1073, 0.0704, 0.0773, 0.1507, 0.1011, 0.0998], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0232, 0.0335, 0.0308, 0.0299, 0.0337, 0.0343, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 05:25:25,798 INFO [train.py:901] (1/4) Epoch 23, batch 3550, loss[loss=0.171, simple_loss=0.2605, pruned_loss=0.04079, over 7253.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2881, pruned_loss=0.06181, over 1614380.23 frames. ], batch size: 16, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:25:39,016 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.351e+02 2.882e+02 3.469e+02 9.271e+02, threshold=5.765e+02, percent-clipped=2.0 +2023-02-07 05:26:01,200 INFO [train.py:901] (1/4) Epoch 23, batch 3600, loss[loss=0.1954, simple_loss=0.2837, pruned_loss=0.05358, over 7643.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2872, pruned_loss=0.06175, over 1612932.22 frames. ], batch size: 19, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:26:30,512 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181465.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:26:37,787 INFO [train.py:901] (1/4) Epoch 23, batch 3650, loss[loss=0.1781, simple_loss=0.2546, pruned_loss=0.0508, over 7527.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2869, pruned_loss=0.06174, over 1605183.29 frames. ], batch size: 18, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:26:48,288 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181490.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:26:50,921 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.424e+02 2.919e+02 3.720e+02 6.119e+02, threshold=5.839e+02, percent-clipped=1.0 +2023-02-07 05:27:11,146 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 05:27:12,455 INFO [train.py:901] (1/4) Epoch 23, batch 3700, loss[loss=0.2203, simple_loss=0.3053, pruned_loss=0.06763, over 8108.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2888, pruned_loss=0.06246, over 1607629.78 frames. ], batch size: 23, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:27:30,282 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181548.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:27:37,416 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181558.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:27:49,548 INFO [train.py:901] (1/4) Epoch 23, batch 3750, loss[loss=0.1703, simple_loss=0.2561, pruned_loss=0.04224, over 7816.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.287, pruned_loss=0.06171, over 1605789.43 frames. ], batch size: 20, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:27:55,389 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181583.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:28:02,824 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 2.354e+02 2.844e+02 3.677e+02 7.170e+02, threshold=5.688e+02, percent-clipped=4.0 +2023-02-07 05:28:24,874 INFO [train.py:901] (1/4) Epoch 23, batch 3800, loss[loss=0.2616, simple_loss=0.325, pruned_loss=0.09911, over 7091.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2857, pruned_loss=0.06101, over 1601753.41 frames. ], batch size: 71, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:28:49,195 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=181659.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:29:00,794 INFO [train.py:901] (1/4) Epoch 23, batch 3850, loss[loss=0.2081, simple_loss=0.3018, pruned_loss=0.05719, over 8319.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2859, pruned_loss=0.061, over 1599656.10 frames. ], batch size: 26, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:29:14,851 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 2.361e+02 2.900e+02 3.650e+02 9.007e+02, threshold=5.800e+02, percent-clipped=7.0 +2023-02-07 05:29:22,382 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 05:29:25,354 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2212, 3.5709, 2.0872, 3.0323, 2.7774, 1.7247, 2.8326, 3.1895], + device='cuda:1'), covar=tensor([0.1726, 0.0440, 0.1441, 0.0761, 0.0868, 0.2064, 0.1287, 0.1079], + device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0232, 0.0337, 0.0310, 0.0300, 0.0338, 0.0343, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 05:29:34,173 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0397, 1.8269, 2.3383, 2.0465, 2.2396, 2.1222, 1.8820, 1.1147], + device='cuda:1'), covar=tensor([0.5199, 0.4334, 0.1792, 0.2976, 0.2086, 0.2690, 0.1839, 0.4637], + device='cuda:1'), in_proj_covar=tensor([0.0949, 0.0995, 0.0816, 0.0957, 0.1004, 0.0906, 0.0758, 0.0834], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 05:29:36,630 INFO [train.py:901] (1/4) Epoch 23, batch 3900, loss[loss=0.2092, simple_loss=0.2936, pruned_loss=0.0624, over 8097.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2866, pruned_loss=0.06117, over 1605010.42 frames. ], batch size: 23, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:30:06,235 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5939, 4.5900, 4.1627, 2.2741, 4.0296, 4.2370, 4.1939, 4.0889], + device='cuda:1'), covar=tensor([0.0675, 0.0501, 0.0902, 0.4605, 0.0836, 0.0941, 0.1089, 0.0645], + device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0441, 0.0430, 0.0540, 0.0429, 0.0444, 0.0428, 0.0388], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:30:07,660 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6652, 1.4565, 2.7992, 1.4431, 2.1895, 3.0333, 3.1721, 2.6004], + device='cuda:1'), covar=tensor([0.1234, 0.1722, 0.0413, 0.2018, 0.0919, 0.0328, 0.0675, 0.0579], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0321, 0.0286, 0.0315, 0.0310, 0.0268, 0.0422, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 05:30:10,464 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181774.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:30:10,998 INFO [train.py:901] (1/4) Epoch 23, batch 3950, loss[loss=0.2109, simple_loss=0.3005, pruned_loss=0.06065, over 8530.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2858, pruned_loss=0.06103, over 1606715.35 frames. ], batch size: 28, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:30:26,257 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.359e+02 2.788e+02 3.393e+02 6.824e+02, threshold=5.575e+02, percent-clipped=4.0 +2023-02-07 05:30:47,699 INFO [train.py:901] (1/4) Epoch 23, batch 4000, loss[loss=0.1889, simple_loss=0.2808, pruned_loss=0.04847, over 8455.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2873, pruned_loss=0.06141, over 1613134.73 frames. ], batch size: 29, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:31:22,582 INFO [train.py:901] (1/4) Epoch 23, batch 4050, loss[loss=0.1986, simple_loss=0.2805, pruned_loss=0.05839, over 8196.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2866, pruned_loss=0.06075, over 1611244.06 frames. ], batch size: 23, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:31:26,927 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5871, 1.6254, 2.1247, 1.3892, 1.3392, 2.0868, 0.3907, 1.2913], + device='cuda:1'), covar=tensor([0.1834, 0.1299, 0.0418, 0.1137, 0.2437, 0.0411, 0.1872, 0.1276], + device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0200, 0.0129, 0.0220, 0.0270, 0.0137, 0.0169, 0.0192], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 05:31:34,364 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=181892.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:31:35,714 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.508e+02 2.885e+02 3.954e+02 8.020e+02, threshold=5.770e+02, percent-clipped=6.0 +2023-02-07 05:31:59,841 INFO [train.py:901] (1/4) Epoch 23, batch 4100, loss[loss=0.1921, simple_loss=0.2679, pruned_loss=0.05814, over 8089.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2875, pruned_loss=0.06137, over 1616079.49 frames. ], batch size: 21, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:32:31,903 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3377, 2.3333, 1.6106, 2.1035, 1.8915, 1.3763, 1.8177, 1.9079], + device='cuda:1'), covar=tensor([0.1666, 0.0422, 0.1421, 0.0672, 0.0786, 0.1709, 0.1114, 0.1060], + device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0235, 0.0339, 0.0311, 0.0302, 0.0339, 0.0346, 0.0319], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 05:32:34,925 INFO [train.py:901] (1/4) Epoch 23, batch 4150, loss[loss=0.1527, simple_loss=0.2323, pruned_loss=0.03657, over 7528.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2878, pruned_loss=0.06125, over 1616099.29 frames. ], batch size: 18, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:32:48,432 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.356e+02 2.929e+02 3.956e+02 6.697e+02, threshold=5.858e+02, percent-clipped=3.0 +2023-02-07 05:32:49,335 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181996.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:32:58,731 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182007.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:33:11,753 INFO [train.py:901] (1/4) Epoch 23, batch 4200, loss[loss=0.2018, simple_loss=0.2812, pruned_loss=0.06117, over 7425.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2875, pruned_loss=0.06133, over 1614447.83 frames. ], batch size: 17, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:33:16,193 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182030.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:33:25,773 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 05:33:33,427 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182055.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:33:47,594 INFO [train.py:901] (1/4) Epoch 23, batch 4250, loss[loss=0.2031, simple_loss=0.2933, pruned_loss=0.05646, over 8505.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2874, pruned_loss=0.06068, over 1619471.14 frames. ], batch size: 28, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:33:49,017 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 05:34:01,333 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.416e+02 2.989e+02 3.588e+02 6.339e+02, threshold=5.979e+02, percent-clipped=2.0 +2023-02-07 05:34:06,319 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5780, 2.4502, 3.1736, 2.4753, 3.1113, 2.5812, 2.3594, 1.8804], + device='cuda:1'), covar=tensor([0.5181, 0.5086, 0.2036, 0.4096, 0.2605, 0.3113, 0.1971, 0.5664], + device='cuda:1'), in_proj_covar=tensor([0.0942, 0.0989, 0.0811, 0.0950, 0.0996, 0.0899, 0.0750, 0.0826], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 05:34:22,741 INFO [train.py:901] (1/4) Epoch 23, batch 4300, loss[loss=0.2053, simple_loss=0.3001, pruned_loss=0.05527, over 8590.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2876, pruned_loss=0.06099, over 1614544.76 frames. ], batch size: 39, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:34:25,780 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8207, 1.5598, 3.3471, 1.4006, 2.4920, 3.6855, 3.8281, 3.1344], + device='cuda:1'), covar=tensor([0.1285, 0.1868, 0.0370, 0.2175, 0.1003, 0.0242, 0.0503, 0.0585], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0320, 0.0286, 0.0315, 0.0311, 0.0268, 0.0422, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 05:34:57,816 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.38 vs. limit=5.0 +2023-02-07 05:34:58,715 INFO [train.py:901] (1/4) Epoch 23, batch 4350, loss[loss=0.1809, simple_loss=0.2608, pruned_loss=0.0505, over 7809.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2872, pruned_loss=0.06074, over 1613243.23 frames. ], batch size: 20, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:35:13,489 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.346e+02 2.960e+02 3.931e+02 9.702e+02, threshold=5.919e+02, percent-clipped=9.0 +2023-02-07 05:35:21,961 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 05:35:34,686 INFO [train.py:901] (1/4) Epoch 23, batch 4400, loss[loss=0.1926, simple_loss=0.2817, pruned_loss=0.0517, over 8328.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2868, pruned_loss=0.06047, over 1610501.15 frames. ], batch size: 25, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:35:46,106 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8335, 1.9702, 1.7674, 2.4912, 1.2767, 1.5192, 1.9156, 2.0310], + device='cuda:1'), covar=tensor([0.0754, 0.0770, 0.0884, 0.0428, 0.1055, 0.1280, 0.0746, 0.0758], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0199, 0.0247, 0.0215, 0.0208, 0.0249, 0.0252, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 05:35:49,489 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7660, 1.5343, 5.8988, 2.1540, 5.2567, 4.9360, 5.4342, 5.3374], + device='cuda:1'), covar=tensor([0.0468, 0.5373, 0.0346, 0.4134, 0.1007, 0.0859, 0.0479, 0.0516], + device='cuda:1'), in_proj_covar=tensor([0.0641, 0.0652, 0.0709, 0.0642, 0.0719, 0.0615, 0.0614, 0.0688], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:36:03,262 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182263.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:36:05,080 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 05:36:11,449 INFO [train.py:901] (1/4) Epoch 23, batch 4450, loss[loss=0.2072, simple_loss=0.2823, pruned_loss=0.06604, over 8354.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2873, pruned_loss=0.06136, over 1610019.89 frames. ], batch size: 26, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:36:20,486 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182288.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:36:26,032 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.606e+02 3.225e+02 4.349e+02 9.132e+02, threshold=6.449e+02, percent-clipped=7.0 +2023-02-07 05:36:28,977 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182299.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:36:44,236 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-07 05:36:47,059 INFO [train.py:901] (1/4) Epoch 23, batch 4500, loss[loss=0.2118, simple_loss=0.2934, pruned_loss=0.0651, over 6819.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2876, pruned_loss=0.06127, over 1611552.67 frames. ], batch size: 15, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:36:54,401 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 05:36:56,811 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 05:36:57,582 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=182340.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:37:16,311 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0052, 2.3616, 1.9434, 3.0066, 1.5857, 1.7021, 2.3280, 2.4403], + device='cuda:1'), covar=tensor([0.0736, 0.0724, 0.0825, 0.0309, 0.1011, 0.1237, 0.0741, 0.0733], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0199, 0.0247, 0.0216, 0.0208, 0.0249, 0.0252, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 05:37:23,617 INFO [train.py:901] (1/4) Epoch 23, batch 4550, loss[loss=0.2591, simple_loss=0.3418, pruned_loss=0.08817, over 8110.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.288, pruned_loss=0.06123, over 1617070.62 frames. ], batch size: 23, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:37:34,406 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-07 05:37:37,490 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.347e+02 2.810e+02 3.651e+02 9.685e+02, threshold=5.619e+02, percent-clipped=2.0 +2023-02-07 05:37:59,265 INFO [train.py:901] (1/4) Epoch 23, batch 4600, loss[loss=0.1957, simple_loss=0.2854, pruned_loss=0.05299, over 8553.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2871, pruned_loss=0.06067, over 1612339.69 frames. ], batch size: 31, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:38:20,349 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182455.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:38:34,844 INFO [train.py:901] (1/4) Epoch 23, batch 4650, loss[loss=0.2049, simple_loss=0.2968, pruned_loss=0.05649, over 8443.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2856, pruned_loss=0.06042, over 1609604.87 frames. ], batch size: 27, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:38:49,709 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-07 05:38:50,627 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.204e+02 2.647e+02 3.638e+02 6.712e+02, threshold=5.294e+02, percent-clipped=7.0 +2023-02-07 05:38:54,471 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 05:39:12,429 INFO [train.py:901] (1/4) Epoch 23, batch 4700, loss[loss=0.1887, simple_loss=0.2726, pruned_loss=0.05238, over 8464.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2865, pruned_loss=0.06078, over 1612816.67 frames. ], batch size: 49, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:39:17,386 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182532.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:39:47,032 INFO [train.py:901] (1/4) Epoch 23, batch 4750, loss[loss=0.163, simple_loss=0.2483, pruned_loss=0.0389, over 7656.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2855, pruned_loss=0.0601, over 1617656.73 frames. ], batch size: 19, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:40:01,599 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.297e+02 2.902e+02 3.418e+02 7.225e+02, threshold=5.805e+02, percent-clipped=3.0 +2023-02-07 05:40:05,946 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 05:40:08,834 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 05:40:24,093 INFO [train.py:901] (1/4) Epoch 23, batch 4800, loss[loss=0.2019, simple_loss=0.277, pruned_loss=0.06335, over 7925.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2847, pruned_loss=0.05958, over 1617298.60 frames. ], batch size: 20, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:40:36,434 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=182643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:40:59,178 INFO [train.py:901] (1/4) Epoch 23, batch 4850, loss[loss=0.2018, simple_loss=0.2936, pruned_loss=0.05504, over 8698.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2848, pruned_loss=0.05943, over 1612818.34 frames. ], batch size: 39, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:41:00,621 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 05:41:13,255 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.288e+02 2.781e+02 3.814e+02 7.165e+02, threshold=5.562e+02, percent-clipped=4.0 +2023-02-07 05:41:25,673 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182711.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:41:36,253 INFO [train.py:901] (1/4) Epoch 23, batch 4900, loss[loss=0.2117, simple_loss=0.3028, pruned_loss=0.06028, over 8102.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2852, pruned_loss=0.05986, over 1615300.33 frames. ], batch size: 23, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:41:40,414 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6260, 5.7300, 5.0325, 2.3131, 5.0346, 5.3626, 5.1865, 5.2411], + device='cuda:1'), covar=tensor([0.0554, 0.0358, 0.0819, 0.4773, 0.0661, 0.0651, 0.1033, 0.0539], + device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0438, 0.0426, 0.0535, 0.0424, 0.0442, 0.0425, 0.0384], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:41:44,975 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182736.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:42:00,357 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182758.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:42:12,971 INFO [train.py:901] (1/4) Epoch 23, batch 4950, loss[loss=0.1809, simple_loss=0.2619, pruned_loss=0.04997, over 5964.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2847, pruned_loss=0.05937, over 1614133.51 frames. ], batch size: 13, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:42:27,044 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.423e+02 2.989e+02 3.745e+02 1.524e+03, threshold=5.977e+02, percent-clipped=7.0 +2023-02-07 05:42:48,229 INFO [train.py:901] (1/4) Epoch 23, batch 5000, loss[loss=0.2148, simple_loss=0.2969, pruned_loss=0.06633, over 8501.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2841, pruned_loss=0.05938, over 1614824.89 frames. ], batch size: 26, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:43:13,253 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9065, 1.5999, 3.4970, 1.5482, 2.4915, 3.9205, 3.9709, 3.3557], + device='cuda:1'), covar=tensor([0.1162, 0.1695, 0.0318, 0.1969, 0.1069, 0.0203, 0.0420, 0.0491], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0320, 0.0286, 0.0315, 0.0312, 0.0268, 0.0423, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 05:43:25,214 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182874.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:43:25,712 INFO [train.py:901] (1/4) Epoch 23, batch 5050, loss[loss=0.1968, simple_loss=0.2574, pruned_loss=0.06814, over 7423.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2849, pruned_loss=0.05995, over 1615431.56 frames. ], batch size: 17, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:43:26,530 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=182876.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:43:29,515 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182880.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:43:40,728 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.376e+02 2.932e+02 3.646e+02 6.966e+02, threshold=5.864e+02, percent-clipped=3.0 +2023-02-07 05:43:46,359 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 05:43:57,144 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2220, 1.9110, 2.5291, 2.1668, 2.5196, 2.2156, 2.0310, 1.3430], + device='cuda:1'), covar=tensor([0.5403, 0.4861, 0.2019, 0.3334, 0.2141, 0.2907, 0.1867, 0.5034], + device='cuda:1'), in_proj_covar=tensor([0.0943, 0.0989, 0.0809, 0.0951, 0.0997, 0.0900, 0.0752, 0.0831], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 05:44:01,709 INFO [train.py:901] (1/4) Epoch 23, batch 5100, loss[loss=0.1555, simple_loss=0.2306, pruned_loss=0.04017, over 7703.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2847, pruned_loss=0.0596, over 1615574.91 frames. ], batch size: 18, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:44:03,387 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 05:44:38,859 INFO [train.py:901] (1/4) Epoch 23, batch 5150, loss[loss=0.225, simple_loss=0.3095, pruned_loss=0.0702, over 8621.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2849, pruned_loss=0.06006, over 1617445.16 frames. ], batch size: 39, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:44:50,225 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182991.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:44:53,605 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.308e+02 2.409e+02 2.843e+02 3.449e+02 6.604e+02, threshold=5.686e+02, percent-clipped=1.0 +2023-02-07 05:44:57,415 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6078, 1.4211, 1.5942, 1.3923, 0.8782, 1.4207, 1.4813, 1.3737], + device='cuda:1'), covar=tensor([0.0593, 0.1231, 0.1664, 0.1399, 0.0616, 0.1440, 0.0725, 0.0645], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0189, 0.0160, 0.0101, 0.0162, 0.0112, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 05:45:03,992 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-07 05:45:07,243 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183014.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:45:14,581 INFO [train.py:901] (1/4) Epoch 23, batch 5200, loss[loss=0.2486, simple_loss=0.3278, pruned_loss=0.08466, over 8100.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2859, pruned_loss=0.06056, over 1615151.69 frames. ], batch size: 23, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:45:24,446 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183039.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:45:46,583 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 05:45:50,598 INFO [train.py:901] (1/4) Epoch 23, batch 5250, loss[loss=0.2334, simple_loss=0.3089, pruned_loss=0.07893, over 8197.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2865, pruned_loss=0.06124, over 1612956.23 frames. ], batch size: 23, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:45:53,706 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.78 vs. limit=5.0 +2023-02-07 05:46:05,167 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.492e+02 2.942e+02 3.798e+02 7.403e+02, threshold=5.885e+02, percent-clipped=3.0 +2023-02-07 05:46:27,071 INFO [train.py:901] (1/4) Epoch 23, batch 5300, loss[loss=0.2016, simple_loss=0.2766, pruned_loss=0.06337, over 8237.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2863, pruned_loss=0.06081, over 1613212.16 frames. ], batch size: 22, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:46:43,509 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 +2023-02-07 05:47:02,889 INFO [train.py:901] (1/4) Epoch 23, batch 5350, loss[loss=0.2002, simple_loss=0.272, pruned_loss=0.06422, over 7292.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2844, pruned_loss=0.05988, over 1612857.32 frames. ], batch size: 16, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:47:03,829 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4914, 1.4552, 1.8480, 1.2360, 1.1998, 1.8458, 0.2606, 1.1285], + device='cuda:1'), covar=tensor([0.1633, 0.1142, 0.0357, 0.0954, 0.2428, 0.0374, 0.1847, 0.1129], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0195, 0.0127, 0.0216, 0.0266, 0.0134, 0.0167, 0.0190], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 05:47:17,713 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.502e+02 3.193e+02 3.793e+02 7.809e+02, threshold=6.385e+02, percent-clipped=1.0 +2023-02-07 05:47:34,660 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:47:38,840 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=183224.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:47:39,484 INFO [train.py:901] (1/4) Epoch 23, batch 5400, loss[loss=0.1989, simple_loss=0.2895, pruned_loss=0.05411, over 8530.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2857, pruned_loss=0.06082, over 1612790.57 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 4.0 +2023-02-07 05:47:55,892 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183247.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:48:13,010 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183272.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:48:14,972 INFO [train.py:901] (1/4) Epoch 23, batch 5450, loss[loss=0.1735, simple_loss=0.27, pruned_loss=0.03854, over 7986.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2866, pruned_loss=0.06081, over 1619768.11 frames. ], batch size: 21, lr: 3.27e-03, grad_scale: 4.0 +2023-02-07 05:48:15,142 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0931, 1.6122, 4.4462, 2.0774, 2.5881, 5.1154, 5.1486, 4.3618], + device='cuda:1'), covar=tensor([0.1240, 0.1838, 0.0268, 0.1845, 0.1125, 0.0174, 0.0305, 0.0537], + device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0318, 0.0284, 0.0313, 0.0311, 0.0266, 0.0421, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 05:48:17,364 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3955, 1.6473, 2.1791, 1.2767, 1.4916, 1.6661, 1.4975, 1.4781], + device='cuda:1'), covar=tensor([0.1973, 0.2457, 0.0929, 0.4482, 0.1892, 0.3403, 0.2380, 0.2327], + device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0608, 0.0555, 0.0647, 0.0648, 0.0594, 0.0540, 0.0631], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:48:27,836 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4852, 2.7657, 2.1903, 3.7724, 1.4979, 2.0439, 2.3803, 2.7411], + device='cuda:1'), covar=tensor([0.0656, 0.0836, 0.0790, 0.0258, 0.1113, 0.1170, 0.0965, 0.0761], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0196, 0.0244, 0.0213, 0.0205, 0.0244, 0.0249, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 05:48:30,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.334e+02 2.819e+02 3.622e+02 6.725e+02, threshold=5.637e+02, percent-clipped=1.0 +2023-02-07 05:48:41,158 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 05:48:52,618 INFO [train.py:901] (1/4) Epoch 23, batch 5500, loss[loss=0.1879, simple_loss=0.267, pruned_loss=0.05442, over 7919.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2856, pruned_loss=0.06061, over 1614541.74 frames. ], batch size: 20, lr: 3.27e-03, grad_scale: 4.0 +2023-02-07 05:48:58,339 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183333.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:49:02,360 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183339.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:49:27,067 INFO [train.py:901] (1/4) Epoch 23, batch 5550, loss[loss=0.2029, simple_loss=0.2941, pruned_loss=0.05588, over 8107.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2863, pruned_loss=0.06132, over 1615418.64 frames. ], batch size: 23, lr: 3.27e-03, grad_scale: 4.0 +2023-02-07 05:49:41,517 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.428e+02 3.119e+02 4.010e+02 1.058e+03, threshold=6.238e+02, percent-clipped=9.0 +2023-02-07 05:50:03,271 INFO [train.py:901] (1/4) Epoch 23, batch 5600, loss[loss=0.2007, simple_loss=0.3052, pruned_loss=0.04811, over 8473.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.286, pruned_loss=0.0608, over 1614718.53 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:50:04,084 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183426.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:50:39,035 INFO [train.py:901] (1/4) Epoch 23, batch 5650, loss[loss=0.1875, simple_loss=0.2763, pruned_loss=0.04932, over 7654.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2864, pruned_loss=0.0614, over 1609735.66 frames. ], batch size: 19, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:50:51,436 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 05:50:53,299 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.300e+02 3.084e+02 3.921e+02 7.530e+02, threshold=6.168e+02, percent-clipped=4.0 +2023-02-07 05:51:14,120 INFO [train.py:901] (1/4) Epoch 23, batch 5700, loss[loss=0.1897, simple_loss=0.2706, pruned_loss=0.05443, over 8671.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2859, pruned_loss=0.06102, over 1611001.45 frames. ], batch size: 34, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:51:38,825 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1624, 2.2697, 1.8468, 2.7772, 1.3926, 1.7085, 2.1133, 2.3028], + device='cuda:1'), covar=tensor([0.0605, 0.0685, 0.0779, 0.0372, 0.1003, 0.1221, 0.0786, 0.0671], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0198, 0.0247, 0.0215, 0.0208, 0.0248, 0.0252, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 05:51:50,330 INFO [train.py:901] (1/4) Epoch 23, batch 5750, loss[loss=0.2031, simple_loss=0.2806, pruned_loss=0.0628, over 8589.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2874, pruned_loss=0.0616, over 1613910.74 frames. ], batch size: 34, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:51:57,138 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 05:52:00,832 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183589.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:52:03,077 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 05:52:05,021 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183595.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:52:05,469 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.334e+02 3.030e+02 3.740e+02 1.347e+03, threshold=6.060e+02, percent-clipped=7.0 +2023-02-07 05:52:17,390 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7034, 2.0359, 2.1206, 1.2283, 2.2074, 1.4760, 0.7174, 1.9813], + device='cuda:1'), covar=tensor([0.0595, 0.0335, 0.0295, 0.0615, 0.0424, 0.0855, 0.0848, 0.0288], + device='cuda:1'), in_proj_covar=tensor([0.0458, 0.0397, 0.0349, 0.0451, 0.0384, 0.0538, 0.0395, 0.0426], + device='cuda:1'), out_proj_covar=tensor([1.2229e-04, 1.0367e-04, 9.1667e-05, 1.1846e-04, 1.0105e-04, 1.5145e-04, + 1.0654e-04, 1.1239e-04], device='cuda:1') +2023-02-07 05:52:18,051 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183614.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:52:22,089 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183620.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:52:25,347 INFO [train.py:901] (1/4) Epoch 23, batch 5800, loss[loss=0.2121, simple_loss=0.2976, pruned_loss=0.06327, over 8245.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2868, pruned_loss=0.06111, over 1615904.47 frames. ], batch size: 24, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:52:30,169 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183632.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:53:01,092 INFO [train.py:901] (1/4) Epoch 23, batch 5850, loss[loss=0.198, simple_loss=0.2678, pruned_loss=0.06405, over 7978.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2865, pruned_loss=0.06122, over 1613893.21 frames. ], batch size: 21, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:53:16,205 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.232e+02 2.367e+02 2.798e+02 3.640e+02 5.951e+02, threshold=5.597e+02, percent-clipped=0.0 +2023-02-07 05:53:36,759 INFO [train.py:901] (1/4) Epoch 23, batch 5900, loss[loss=0.1876, simple_loss=0.2721, pruned_loss=0.05159, over 8286.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2855, pruned_loss=0.06056, over 1613892.93 frames. ], batch size: 23, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:53:58,150 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0570, 1.6954, 1.8869, 1.5509, 1.0154, 1.6449, 1.8560, 1.8236], + device='cuda:1'), covar=tensor([0.0537, 0.1201, 0.1528, 0.1375, 0.0569, 0.1341, 0.0642, 0.0597], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0188, 0.0159, 0.0100, 0.0161, 0.0112, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:1') +2023-02-07 05:54:08,425 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=183770.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:54:11,757 INFO [train.py:901] (1/4) Epoch 23, batch 5950, loss[loss=0.1913, simple_loss=0.2726, pruned_loss=0.05496, over 7554.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2861, pruned_loss=0.06062, over 1612000.62 frames. ], batch size: 18, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:54:27,013 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.423e+02 2.784e+02 3.423e+02 5.836e+02, threshold=5.567e+02, percent-clipped=2.0 +2023-02-07 05:54:47,616 INFO [train.py:901] (1/4) Epoch 23, batch 6000, loss[loss=0.1727, simple_loss=0.2592, pruned_loss=0.04311, over 7976.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2884, pruned_loss=0.06221, over 1614709.30 frames. ], batch size: 21, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:54:47,616 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 05:55:00,698 INFO [train.py:935] (1/4) Epoch 23, validation: loss=0.1722, simple_loss=0.2724, pruned_loss=0.03597, over 944034.00 frames. +2023-02-07 05:55:00,699 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 05:55:25,807 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183860.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:55:36,139 INFO [train.py:901] (1/4) Epoch 23, batch 6050, loss[loss=0.1966, simple_loss=0.2863, pruned_loss=0.05344, over 8497.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2868, pruned_loss=0.06149, over 1612210.35 frames. ], batch size: 26, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:55:43,238 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183885.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:55:50,620 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.465e+02 3.097e+02 3.782e+02 8.398e+02, threshold=6.194e+02, percent-clipped=6.0 +2023-02-07 05:56:11,855 INFO [train.py:901] (1/4) Epoch 23, batch 6100, loss[loss=0.162, simple_loss=0.2496, pruned_loss=0.03719, over 8145.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2867, pruned_loss=0.06202, over 1609817.36 frames. ], batch size: 22, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:56:21,661 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 05:56:23,561 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8748, 3.8036, 3.5139, 1.8577, 3.4154, 3.4468, 3.4886, 3.3586], + device='cuda:1'), covar=tensor([0.0799, 0.0626, 0.1003, 0.4276, 0.0896, 0.1147, 0.1209, 0.0815], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0443, 0.0431, 0.0539, 0.0430, 0.0446, 0.0428, 0.0386], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:56:32,462 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 05:56:47,365 INFO [train.py:901] (1/4) Epoch 23, batch 6150, loss[loss=0.201, simple_loss=0.2815, pruned_loss=0.0602, over 7654.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2866, pruned_loss=0.06161, over 1610503.63 frames. ], batch size: 19, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:56:48,167 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=183976.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:56:51,715 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9571, 1.6160, 3.4409, 1.6052, 2.4267, 3.7815, 3.9057, 3.2153], + device='cuda:1'), covar=tensor([0.1136, 0.1653, 0.0297, 0.1996, 0.0995, 0.0219, 0.0471, 0.0514], + device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0317, 0.0282, 0.0312, 0.0308, 0.0265, 0.0419, 0.0300], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 05:56:59,934 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5399, 1.5383, 4.7718, 1.8009, 4.1950, 3.9326, 4.3185, 4.2067], + device='cuda:1'), covar=tensor([0.0590, 0.4971, 0.0469, 0.4341, 0.1049, 0.1054, 0.0536, 0.0621], + device='cuda:1'), in_proj_covar=tensor([0.0646, 0.0654, 0.0708, 0.0643, 0.0723, 0.0619, 0.0618, 0.0693], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:57:01,785 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.512e+02 2.876e+02 3.577e+02 6.799e+02, threshold=5.752e+02, percent-clipped=2.0 +2023-02-07 05:57:22,984 INFO [train.py:901] (1/4) Epoch 23, batch 6200, loss[loss=0.1753, simple_loss=0.2574, pruned_loss=0.04659, over 7804.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2863, pruned_loss=0.06171, over 1604405.41 frames. ], batch size: 20, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:57:23,224 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9698, 1.6131, 1.4881, 1.5602, 1.3583, 1.3222, 1.2908, 1.2847], + device='cuda:1'), covar=tensor([0.1315, 0.0502, 0.1362, 0.0629, 0.0851, 0.1641, 0.1023, 0.0850], + device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0231, 0.0332, 0.0306, 0.0298, 0.0338, 0.0341, 0.0315], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 05:57:54,991 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184068.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:57:59,615 INFO [train.py:901] (1/4) Epoch 23, batch 6250, loss[loss=0.2016, simple_loss=0.2868, pruned_loss=0.05821, over 8522.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2856, pruned_loss=0.06055, over 1608443.23 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:58:04,010 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4747, 1.4138, 1.6407, 1.2766, 0.9549, 1.4012, 1.4555, 1.3332], + device='cuda:1'), covar=tensor([0.0633, 0.1263, 0.1619, 0.1497, 0.0597, 0.1510, 0.0749, 0.0696], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0159, 0.0100, 0.0162, 0.0111, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 05:58:06,672 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5495, 1.8781, 2.7451, 1.3736, 2.0625, 1.8467, 1.5765, 1.9601], + device='cuda:1'), covar=tensor([0.2042, 0.2463, 0.0901, 0.4559, 0.1819, 0.3277, 0.2404, 0.2307], + device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0611, 0.0554, 0.0647, 0.0647, 0.0594, 0.0542, 0.0631], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 05:58:11,443 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184091.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:58:14,642 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.476e+02 2.340e+02 2.866e+02 3.425e+02 5.984e+02, threshold=5.731e+02, percent-clipped=3.0 +2023-02-07 05:58:34,480 INFO [train.py:901] (1/4) Epoch 23, batch 6300, loss[loss=0.1601, simple_loss=0.2316, pruned_loss=0.04429, over 7436.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2849, pruned_loss=0.05985, over 1611282.23 frames. ], batch size: 17, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:58:42,583 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.82 vs. limit=5.0 +2023-02-07 05:58:45,749 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184141.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:59:04,544 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184166.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:59:10,625 INFO [train.py:901] (1/4) Epoch 23, batch 6350, loss[loss=0.1966, simple_loss=0.2861, pruned_loss=0.05355, over 8550.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2849, pruned_loss=0.06003, over 1611195.61 frames. ], batch size: 31, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 05:59:25,790 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.298e+02 2.703e+02 3.593e+02 9.198e+02, threshold=5.406e+02, percent-clipped=6.0 +2023-02-07 05:59:32,344 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=184204.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:59:46,835 INFO [train.py:901] (1/4) Epoch 23, batch 6400, loss[loss=0.2092, simple_loss=0.2913, pruned_loss=0.06351, over 8526.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2847, pruned_loss=0.06001, over 1612392.48 frames. ], batch size: 31, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:00:22,071 INFO [train.py:901] (1/4) Epoch 23, batch 6450, loss[loss=0.1711, simple_loss=0.2552, pruned_loss=0.04353, over 7938.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.285, pruned_loss=0.06011, over 1615952.75 frames. ], batch size: 20, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:00:31,217 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.4437, 1.6025, 3.7750, 1.5870, 3.0114, 2.9289, 3.3289, 3.3377], + device='cuda:1'), covar=tensor([0.1761, 0.6541, 0.1602, 0.5833, 0.2655, 0.2401, 0.1505, 0.1365], + device='cuda:1'), in_proj_covar=tensor([0.0642, 0.0649, 0.0706, 0.0640, 0.0716, 0.0616, 0.0616, 0.0691], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:00:37,219 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.429e+02 3.055e+02 3.904e+02 7.071e+02, threshold=6.109e+02, percent-clipped=5.0 +2023-02-07 06:00:54,557 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184319.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:00:58,473 INFO [train.py:901] (1/4) Epoch 23, batch 6500, loss[loss=0.1952, simple_loss=0.2859, pruned_loss=0.05223, over 8322.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2867, pruned_loss=0.06106, over 1618139.84 frames. ], batch size: 25, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:01:01,451 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.59 vs. limit=5.0 +2023-02-07 06:01:13,635 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184347.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:01:30,735 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184372.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:01:32,650 INFO [train.py:901] (1/4) Epoch 23, batch 6550, loss[loss=0.2215, simple_loss=0.3073, pruned_loss=0.0679, over 8349.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2857, pruned_loss=0.0606, over 1616919.92 frames. ], batch size: 26, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:01:34,218 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184377.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:01:48,099 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.251e+02 2.720e+02 3.518e+02 7.175e+02, threshold=5.440e+02, percent-clipped=6.0 +2023-02-07 06:01:51,594 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 06:02:00,074 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=184412.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:02:07,271 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1496, 2.1906, 2.2651, 1.7231, 2.4306, 1.8683, 1.7796, 2.0671], + device='cuda:1'), covar=tensor([0.0574, 0.0384, 0.0269, 0.0579, 0.0393, 0.0570, 0.0653, 0.0397], + device='cuda:1'), in_proj_covar=tensor([0.0455, 0.0396, 0.0349, 0.0448, 0.0381, 0.0536, 0.0392, 0.0425], + device='cuda:1'), out_proj_covar=tensor([1.2144e-04, 1.0353e-04, 9.1555e-05, 1.1783e-04, 1.0023e-04, 1.5082e-04, + 1.0563e-04, 1.1218e-04], device='cuda:1') +2023-02-07 06:02:09,826 INFO [train.py:901] (1/4) Epoch 23, batch 6600, loss[loss=0.1896, simple_loss=0.2697, pruned_loss=0.05479, over 7661.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2869, pruned_loss=0.06078, over 1616437.31 frames. ], batch size: 19, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:02:09,857 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 06:02:20,774 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-07 06:02:42,918 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 06:02:45,212 INFO [train.py:901] (1/4) Epoch 23, batch 6650, loss[loss=0.1949, simple_loss=0.2858, pruned_loss=0.05204, over 8239.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2862, pruned_loss=0.06006, over 1618275.49 frames. ], batch size: 24, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:02:52,260 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9396, 1.6189, 3.3500, 1.5583, 2.3874, 3.6807, 3.8249, 3.1320], + device='cuda:1'), covar=tensor([0.1228, 0.1736, 0.0360, 0.2163, 0.1106, 0.0247, 0.0550, 0.0600], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0320, 0.0286, 0.0315, 0.0313, 0.0269, 0.0425, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:03:00,392 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.187e+02 2.636e+02 3.150e+02 7.164e+02, threshold=5.273e+02, percent-clipped=1.0 +2023-02-07 06:03:06,015 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184504.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:03:21,185 INFO [train.py:901] (1/4) Epoch 23, batch 6700, loss[loss=0.2408, simple_loss=0.3245, pruned_loss=0.07856, over 8189.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.287, pruned_loss=0.06003, over 1623202.07 frames. ], batch size: 23, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:03:22,772 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184527.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:03:56,972 INFO [train.py:901] (1/4) Epoch 23, batch 6750, loss[loss=0.1999, simple_loss=0.291, pruned_loss=0.05444, over 8203.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2869, pruned_loss=0.06013, over 1623108.13 frames. ], batch size: 23, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:03:57,214 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184575.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:04:11,515 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.299e+02 2.705e+02 3.689e+02 1.087e+03, threshold=5.410e+02, percent-clipped=6.0 +2023-02-07 06:04:14,464 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184600.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:04:30,938 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 06:04:32,276 INFO [train.py:901] (1/4) Epoch 23, batch 6800, loss[loss=0.2275, simple_loss=0.3033, pruned_loss=0.0759, over 7658.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2861, pruned_loss=0.06006, over 1618830.86 frames. ], batch size: 19, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:04:55,642 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 06:05:08,931 INFO [train.py:901] (1/4) Epoch 23, batch 6850, loss[loss=0.1811, simple_loss=0.2633, pruned_loss=0.04938, over 7688.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.287, pruned_loss=0.06078, over 1619308.97 frames. ], batch size: 18, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:05:19,305 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 06:05:23,525 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.645e+02 3.100e+02 4.179e+02 7.238e+02, threshold=6.201e+02, percent-clipped=8.0 +2023-02-07 06:05:40,684 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=184721.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:05:43,472 INFO [train.py:901] (1/4) Epoch 23, batch 6900, loss[loss=0.2183, simple_loss=0.2921, pruned_loss=0.07225, over 7972.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2865, pruned_loss=0.06091, over 1618254.29 frames. ], batch size: 21, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:06:21,332 INFO [train.py:901] (1/4) Epoch 23, batch 6950, loss[loss=0.1837, simple_loss=0.2712, pruned_loss=0.04806, over 8248.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2859, pruned_loss=0.06024, over 1616417.43 frames. ], batch size: 22, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:06:27,134 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184783.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:06:29,046 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 06:06:35,999 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 2.334e+02 2.863e+02 3.573e+02 6.345e+02, threshold=5.727e+02, percent-clipped=1.0 +2023-02-07 06:06:38,088 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4702, 4.4613, 3.9944, 1.8160, 4.0032, 3.9553, 4.0221, 3.9160], + device='cuda:1'), covar=tensor([0.0683, 0.0493, 0.1056, 0.4770, 0.0786, 0.1155, 0.1200, 0.0863], + device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0447, 0.0431, 0.0544, 0.0434, 0.0448, 0.0430, 0.0389], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:06:44,107 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.36 vs. limit=5.0 +2023-02-07 06:06:44,525 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184808.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:06:56,220 INFO [train.py:901] (1/4) Epoch 23, batch 7000, loss[loss=0.1968, simple_loss=0.2902, pruned_loss=0.05172, over 8111.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2862, pruned_loss=0.06029, over 1618721.74 frames. ], batch size: 23, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:07:03,938 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184836.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:07:12,024 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=184848.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:07:32,118 INFO [train.py:901] (1/4) Epoch 23, batch 7050, loss[loss=0.2166, simple_loss=0.3022, pruned_loss=0.06551, over 8632.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2852, pruned_loss=0.05985, over 1615352.11 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:07:38,578 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184884.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:07:48,060 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.247e+02 2.854e+02 3.580e+02 1.056e+03, threshold=5.709e+02, percent-clipped=4.0 +2023-02-07 06:07:53,396 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.26 vs. limit=5.0 +2023-02-07 06:08:08,158 INFO [train.py:901] (1/4) Epoch 23, batch 7100, loss[loss=0.1634, simple_loss=0.2439, pruned_loss=0.04149, over 7274.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2846, pruned_loss=0.0593, over 1611506.54 frames. ], batch size: 16, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:08:30,163 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184957.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:08:34,369 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184963.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:08:42,479 INFO [train.py:901] (1/4) Epoch 23, batch 7150, loss[loss=0.2059, simple_loss=0.2738, pruned_loss=0.06904, over 8084.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2855, pruned_loss=0.05978, over 1611989.53 frames. ], batch size: 21, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:08:53,463 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1773, 2.0123, 2.4781, 2.0944, 2.4095, 2.2551, 2.0753, 1.4013], + device='cuda:1'), covar=tensor([0.5206, 0.4400, 0.1881, 0.3592, 0.2610, 0.2833, 0.1848, 0.5048], + device='cuda:1'), in_proj_covar=tensor([0.0942, 0.0991, 0.0806, 0.0950, 0.0997, 0.0896, 0.0753, 0.0829], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 06:08:58,831 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.323e+02 2.664e+02 3.243e+02 7.163e+02, threshold=5.329e+02, percent-clipped=2.0 +2023-02-07 06:09:20,376 INFO [train.py:901] (1/4) Epoch 23, batch 7200, loss[loss=0.1992, simple_loss=0.2875, pruned_loss=0.05544, over 8457.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2855, pruned_loss=0.05968, over 1615210.77 frames. ], batch size: 29, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:09:29,435 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7445, 1.4521, 1.6751, 1.3252, 0.9024, 1.4286, 1.6045, 1.3578], + device='cuda:1'), covar=tensor([0.0574, 0.1332, 0.1718, 0.1552, 0.0627, 0.1568, 0.0736, 0.0714], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0161, 0.0102, 0.0164, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 06:09:35,042 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9967, 1.4921, 3.4512, 1.5348, 2.4145, 3.8086, 3.9003, 3.3196], + device='cuda:1'), covar=tensor([0.1184, 0.1947, 0.0361, 0.2101, 0.1125, 0.0227, 0.0534, 0.0528], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0323, 0.0287, 0.0315, 0.0314, 0.0271, 0.0426, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:09:35,778 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8135, 1.7558, 2.4346, 1.5608, 1.3094, 2.3993, 0.4241, 1.4573], + device='cuda:1'), covar=tensor([0.1598, 0.1205, 0.0350, 0.1331, 0.2780, 0.0416, 0.2165, 0.1546], + device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0195, 0.0129, 0.0219, 0.0268, 0.0136, 0.0168, 0.0191], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 06:09:44,880 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9490, 1.5007, 1.7956, 1.3050, 1.0488, 1.4556, 1.8353, 1.5230], + device='cuda:1'), covar=tensor([0.0523, 0.1196, 0.1634, 0.1471, 0.0590, 0.1495, 0.0651, 0.0671], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0160, 0.0101, 0.0164, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 06:09:44,924 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7137, 1.6993, 2.3014, 1.5079, 1.2196, 2.2867, 0.4099, 1.4273], + device='cuda:1'), covar=tensor([0.1628, 0.1173, 0.0376, 0.1188, 0.2851, 0.0365, 0.1966, 0.1424], + device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0196, 0.0129, 0.0220, 0.0268, 0.0136, 0.0169, 0.0191], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 06:09:54,644 INFO [train.py:901] (1/4) Epoch 23, batch 7250, loss[loss=0.2406, simple_loss=0.3258, pruned_loss=0.07773, over 8502.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2856, pruned_loss=0.05985, over 1614865.45 frames. ], batch size: 26, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:09:56,904 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185078.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:10:06,366 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185092.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:10:09,704 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.341e+02 2.701e+02 3.625e+02 6.528e+02, threshold=5.401e+02, percent-clipped=8.0 +2023-02-07 06:10:25,009 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185117.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:10:30,397 INFO [train.py:901] (1/4) Epoch 23, batch 7300, loss[loss=0.1876, simple_loss=0.2812, pruned_loss=0.04704, over 8251.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2855, pruned_loss=0.06008, over 1610375.54 frames. ], batch size: 24, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:10:45,741 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5230, 2.3684, 3.0993, 2.4997, 2.9458, 2.5377, 2.3520, 1.9366], + device='cuda:1'), covar=tensor([0.5090, 0.4786, 0.2026, 0.3755, 0.2522, 0.2778, 0.1749, 0.5136], + device='cuda:1'), in_proj_covar=tensor([0.0939, 0.0986, 0.0805, 0.0948, 0.0995, 0.0896, 0.0751, 0.0826], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 06:10:46,962 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185147.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:11:04,728 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.08 vs. limit=5.0 +2023-02-07 06:11:06,505 INFO [train.py:901] (1/4) Epoch 23, batch 7350, loss[loss=0.2055, simple_loss=0.2899, pruned_loss=0.06057, over 8635.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2858, pruned_loss=0.05991, over 1616735.45 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:11:19,727 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 06:11:21,058 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.380e+02 2.863e+02 3.556e+02 7.708e+02, threshold=5.726e+02, percent-clipped=6.0 +2023-02-07 06:11:38,637 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6533, 1.6886, 2.3991, 1.5299, 1.2162, 2.3436, 0.5158, 1.4442], + device='cuda:1'), covar=tensor([0.1707, 0.1274, 0.0344, 0.1312, 0.2808, 0.0412, 0.2041, 0.1340], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0195, 0.0128, 0.0219, 0.0268, 0.0136, 0.0168, 0.0191], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 06:11:38,642 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185219.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:11:41,242 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 06:11:42,633 INFO [train.py:901] (1/4) Epoch 23, batch 7400, loss[loss=0.175, simple_loss=0.2569, pruned_loss=0.04661, over 8085.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2863, pruned_loss=0.06027, over 1616151.81 frames. ], batch size: 21, lr: 3.26e-03, grad_scale: 16.0 +2023-02-07 06:11:44,813 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185228.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:11:48,317 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6578, 1.5113, 3.0894, 1.4970, 2.2216, 3.3590, 3.5193, 2.8467], + device='cuda:1'), covar=tensor([0.1383, 0.1849, 0.0369, 0.2107, 0.1044, 0.0277, 0.0500, 0.0583], + device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0322, 0.0287, 0.0315, 0.0314, 0.0269, 0.0424, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:11:56,709 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185244.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:12:04,423 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-07 06:12:18,690 INFO [train.py:901] (1/4) Epoch 23, batch 7450, loss[loss=0.1622, simple_loss=0.2408, pruned_loss=0.04181, over 7412.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2856, pruned_loss=0.0599, over 1615747.14 frames. ], batch size: 17, lr: 3.26e-03, grad_scale: 16.0 +2023-02-07 06:12:18,941 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4540, 1.8399, 2.6113, 1.3287, 1.9419, 1.8182, 1.5751, 1.9613], + device='cuda:1'), covar=tensor([0.2072, 0.2595, 0.0968, 0.4695, 0.2018, 0.3479, 0.2423, 0.2358], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0611, 0.0554, 0.0646, 0.0647, 0.0595, 0.0541, 0.0632], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:12:21,573 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 06:12:33,469 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.310e+02 2.954e+02 3.827e+02 6.869e+02, threshold=5.908e+02, percent-clipped=4.0 +2023-02-07 06:12:37,076 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185301.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:12:53,819 INFO [train.py:901] (1/4) Epoch 23, batch 7500, loss[loss=0.19, simple_loss=0.2841, pruned_loss=0.04794, over 8770.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2854, pruned_loss=0.05984, over 1614843.10 frames. ], batch size: 30, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:13:05,839 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-07 06:13:08,257 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185343.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:13:22,311 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7490, 1.9051, 1.6520, 2.3197, 0.9102, 1.4426, 1.6913, 1.8536], + device='cuda:1'), covar=tensor([0.0748, 0.0801, 0.0925, 0.0455, 0.1164, 0.1342, 0.0767, 0.0722], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0199, 0.0244, 0.0214, 0.0206, 0.0247, 0.0250, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 06:13:31,741 INFO [train.py:901] (1/4) Epoch 23, batch 7550, loss[loss=0.2063, simple_loss=0.2948, pruned_loss=0.05887, over 8135.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.284, pruned_loss=0.05922, over 1610059.33 frames. ], batch size: 22, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:13:39,222 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 +2023-02-07 06:13:46,072 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.461e+02 3.059e+02 3.860e+02 7.244e+02, threshold=6.118e+02, percent-clipped=3.0 +2023-02-07 06:14:00,488 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185416.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:14:04,572 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185422.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:14:06,594 INFO [train.py:901] (1/4) Epoch 23, batch 7600, loss[loss=0.1907, simple_loss=0.2864, pruned_loss=0.0475, over 8528.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2854, pruned_loss=0.05997, over 1614843.32 frames. ], batch size: 28, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:14:09,580 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.82 vs. limit=5.0 +2023-02-07 06:14:41,950 INFO [train.py:901] (1/4) Epoch 23, batch 7650, loss[loss=0.2261, simple_loss=0.3109, pruned_loss=0.07067, over 8187.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2875, pruned_loss=0.06126, over 1612492.98 frames. ], batch size: 23, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:14:50,197 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185486.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:14:54,393 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185491.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:14:57,808 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.478e+02 3.100e+02 3.999e+02 8.387e+02, threshold=6.200e+02, percent-clipped=6.0 +2023-02-07 06:15:17,431 INFO [train.py:901] (1/4) Epoch 23, batch 7700, loss[loss=0.1932, simple_loss=0.2853, pruned_loss=0.05052, over 8484.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2879, pruned_loss=0.06161, over 1614391.73 frames. ], batch size: 26, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:15:25,723 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185537.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:15:37,332 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 06:15:53,097 INFO [train.py:901] (1/4) Epoch 23, batch 7750, loss[loss=0.1728, simple_loss=0.2469, pruned_loss=0.04935, over 7234.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.287, pruned_loss=0.06136, over 1613260.83 frames. ], batch size: 16, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:16:08,174 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.336e+02 2.905e+02 3.607e+02 6.527e+02, threshold=5.810e+02, percent-clipped=2.0 +2023-02-07 06:16:10,501 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185599.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:16:15,285 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185606.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:16:28,328 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185624.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:16:28,832 INFO [train.py:901] (1/4) Epoch 23, batch 7800, loss[loss=0.1781, simple_loss=0.2598, pruned_loss=0.04818, over 7632.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2879, pruned_loss=0.06148, over 1614276.97 frames. ], batch size: 19, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:17:01,061 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185672.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:17:02,897 INFO [train.py:901] (1/4) Epoch 23, batch 7850, loss[loss=0.1878, simple_loss=0.277, pruned_loss=0.04929, over 8598.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2867, pruned_loss=0.06068, over 1615966.77 frames. ], batch size: 39, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:17:17,284 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.434e+02 2.983e+02 3.607e+02 9.941e+02, threshold=5.966e+02, percent-clipped=5.0 +2023-02-07 06:17:18,191 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185697.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:17:37,220 INFO [train.py:901] (1/4) Epoch 23, batch 7900, loss[loss=0.1766, simple_loss=0.2617, pruned_loss=0.04569, over 8249.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2864, pruned_loss=0.06074, over 1616953.20 frames. ], batch size: 22, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:18:00,629 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-07 06:18:11,079 INFO [train.py:901] (1/4) Epoch 23, batch 7950, loss[loss=0.1852, simple_loss=0.2641, pruned_loss=0.05313, over 7937.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.286, pruned_loss=0.06067, over 1614272.16 frames. ], batch size: 20, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:18:12,602 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185777.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:18:12,816 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 +2023-02-07 06:18:23,319 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185793.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:18:25,078 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 2.254e+02 2.775e+02 3.427e+02 8.244e+02, threshold=5.550e+02, percent-clipped=2.0 +2023-02-07 06:18:39,377 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185817.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 06:18:40,124 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185818.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:18:44,666 INFO [train.py:901] (1/4) Epoch 23, batch 8000, loss[loss=0.1813, simple_loss=0.2597, pruned_loss=0.0514, over 7243.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2865, pruned_loss=0.06111, over 1611184.98 frames. ], batch size: 16, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:18:48,042 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185830.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:19:09,858 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185862.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:19:18,259 INFO [train.py:901] (1/4) Epoch 23, batch 8050, loss[loss=0.1906, simple_loss=0.2659, pruned_loss=0.0577, over 7186.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2847, pruned_loss=0.06101, over 1595863.28 frames. ], batch size: 16, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:19:26,748 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185887.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:19:32,782 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.662e+02 3.318e+02 4.159e+02 9.358e+02, threshold=6.635e+02, percent-clipped=7.0 +2023-02-07 06:19:52,111 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 06:19:57,779 INFO [train.py:901] (1/4) Epoch 24, batch 0, loss[loss=0.1897, simple_loss=0.2687, pruned_loss=0.05531, over 7649.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2687, pruned_loss=0.05531, over 7649.00 frames. ], batch size: 19, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:19:57,779 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 06:20:01,759 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6080, 1.3301, 1.6010, 1.3133, 0.9141, 1.3521, 1.5892, 1.1963], + device='cuda:1'), covar=tensor([0.0662, 0.1378, 0.1785, 0.1571, 0.0646, 0.1592, 0.0714, 0.0755], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0160, 0.0101, 0.0164, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 06:20:09,069 INFO [train.py:935] (1/4) Epoch 24, validation: loss=0.1731, simple_loss=0.2733, pruned_loss=0.03644, over 944034.00 frames. +2023-02-07 06:20:09,070 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 06:20:23,906 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 06:20:35,523 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185945.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:20:44,023 INFO [train.py:901] (1/4) Epoch 24, batch 50, loss[loss=0.2098, simple_loss=0.2839, pruned_loss=0.06782, over 7654.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2939, pruned_loss=0.06255, over 370899.98 frames. ], batch size: 19, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:20:57,549 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 06:21:11,398 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.437e+02 2.851e+02 3.663e+02 1.155e+03, threshold=5.702e+02, percent-clipped=3.0 +2023-02-07 06:21:20,552 INFO [train.py:901] (1/4) Epoch 24, batch 100, loss[loss=0.2199, simple_loss=0.3021, pruned_loss=0.06883, over 8349.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2916, pruned_loss=0.06198, over 649948.09 frames. ], batch size: 24, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:21:22,592 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 06:21:41,726 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4151, 1.4028, 4.5939, 1.7490, 4.1582, 3.8287, 4.2037, 4.0903], + device='cuda:1'), covar=tensor([0.0499, 0.4783, 0.0467, 0.3939, 0.0895, 0.0935, 0.0500, 0.0593], + device='cuda:1'), in_proj_covar=tensor([0.0645, 0.0655, 0.0708, 0.0641, 0.0720, 0.0620, 0.0615, 0.0690], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:21:56,098 INFO [train.py:901] (1/4) Epoch 24, batch 150, loss[loss=0.1889, simple_loss=0.2749, pruned_loss=0.05145, over 8599.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2858, pruned_loss=0.05925, over 864495.97 frames. ], batch size: 39, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:22:00,569 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.61 vs. limit=5.0 +2023-02-07 06:22:03,342 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-07 06:22:07,742 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8248, 2.0346, 2.2324, 1.4549, 2.3485, 1.6933, 0.6974, 2.0159], + device='cuda:1'), covar=tensor([0.0645, 0.0396, 0.0300, 0.0624, 0.0441, 0.0836, 0.0976, 0.0321], + device='cuda:1'), in_proj_covar=tensor([0.0462, 0.0402, 0.0354, 0.0455, 0.0385, 0.0545, 0.0398, 0.0428], + device='cuda:1'), out_proj_covar=tensor([1.2331e-04, 1.0515e-04, 9.2994e-05, 1.1942e-04, 1.0126e-04, 1.5330e-04, + 1.0710e-04, 1.1294e-04], device='cuda:1') +2023-02-07 06:22:21,897 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.384e+02 2.880e+02 3.401e+02 7.597e+02, threshold=5.761e+02, percent-clipped=1.0 +2023-02-07 06:22:30,264 INFO [train.py:901] (1/4) Epoch 24, batch 200, loss[loss=0.1902, simple_loss=0.2838, pruned_loss=0.04834, over 8551.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2854, pruned_loss=0.05927, over 1028505.80 frames. ], batch size: 31, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:22:35,401 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4164, 2.0902, 3.1604, 2.0614, 2.7574, 3.5444, 3.4468, 3.2266], + device='cuda:1'), covar=tensor([0.0896, 0.1479, 0.0564, 0.1750, 0.1489, 0.0218, 0.0595, 0.0416], + device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0321, 0.0286, 0.0317, 0.0314, 0.0269, 0.0426, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:22:38,161 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:22:40,033 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186121.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:23:01,364 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-07 06:23:05,569 INFO [train.py:901] (1/4) Epoch 24, batch 250, loss[loss=0.2393, simple_loss=0.315, pruned_loss=0.08182, over 8242.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2854, pruned_loss=0.05915, over 1163397.34 frames. ], batch size: 24, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:23:07,696 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186161.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 06:23:16,524 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 06:23:18,803 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186176.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:23:25,615 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 06:23:32,345 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.402e+02 3.098e+02 3.972e+02 8.418e+02, threshold=6.197e+02, percent-clipped=5.0 +2023-02-07 06:23:36,038 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186201.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:23:40,511 INFO [train.py:901] (1/4) Epoch 24, batch 300, loss[loss=0.2042, simple_loss=0.2861, pruned_loss=0.06112, over 8774.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2852, pruned_loss=0.05912, over 1263739.60 frames. ], batch size: 30, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:23:53,001 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186226.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:24:00,544 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186236.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:24:00,862 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-07 06:24:13,919 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6153, 2.6918, 2.0325, 2.5115, 2.2200, 1.8442, 2.2762, 2.2916], + device='cuda:1'), covar=tensor([0.1539, 0.0423, 0.1191, 0.0600, 0.0773, 0.1454, 0.0995, 0.0979], + device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0235, 0.0337, 0.0311, 0.0303, 0.0342, 0.0350, 0.0320], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 06:24:15,080 INFO [train.py:901] (1/4) Epoch 24, batch 350, loss[loss=0.2069, simple_loss=0.2975, pruned_loss=0.0581, over 8451.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2844, pruned_loss=0.0589, over 1342667.08 frames. ], batch size: 25, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:24:24,623 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6764, 4.6407, 4.1935, 2.0638, 4.1422, 4.2792, 4.2356, 4.1756], + device='cuda:1'), covar=tensor([0.0635, 0.0506, 0.0926, 0.4400, 0.0785, 0.0935, 0.1076, 0.0737], + device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0445, 0.0431, 0.0541, 0.0431, 0.0446, 0.0426, 0.0389], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:24:28,144 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186276.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 06:24:28,714 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186277.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:24:42,181 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.476e+02 2.428e+02 2.971e+02 3.348e+02 5.777e+02, threshold=5.941e+02, percent-clipped=0.0 +2023-02-07 06:24:47,691 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.3111, 3.1805, 2.9791, 1.6594, 2.9085, 2.9583, 2.8925, 2.8177], + device='cuda:1'), covar=tensor([0.1050, 0.0766, 0.1289, 0.4198, 0.1167, 0.1277, 0.1500, 0.1144], + device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0445, 0.0431, 0.0541, 0.0431, 0.0446, 0.0426, 0.0389], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:24:50,336 INFO [train.py:901] (1/4) Epoch 24, batch 400, loss[loss=0.1994, simple_loss=0.2641, pruned_loss=0.06733, over 8113.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2853, pruned_loss=0.05955, over 1404274.69 frames. ], batch size: 23, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:25:26,091 INFO [train.py:901] (1/4) Epoch 24, batch 450, loss[loss=0.2029, simple_loss=0.2915, pruned_loss=0.05708, over 8509.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2863, pruned_loss=0.06026, over 1452744.95 frames. ], batch size: 26, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:25:52,934 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.487e+02 2.919e+02 3.580e+02 7.824e+02, threshold=5.839e+02, percent-clipped=3.0 +2023-02-07 06:26:02,029 INFO [train.py:901] (1/4) Epoch 24, batch 500, loss[loss=0.2174, simple_loss=0.3034, pruned_loss=0.06575, over 8511.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2862, pruned_loss=0.06024, over 1489494.38 frames. ], batch size: 28, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:26:23,342 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186439.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:26:37,816 INFO [train.py:901] (1/4) Epoch 24, batch 550, loss[loss=0.2164, simple_loss=0.3047, pruned_loss=0.0641, over 8329.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2863, pruned_loss=0.06041, over 1518052.86 frames. ], batch size: 26, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:26:40,765 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186462.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:26:55,561 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7653, 1.9327, 2.1133, 1.2759, 2.2192, 1.6391, 0.6881, 1.9832], + device='cuda:1'), covar=tensor([0.0652, 0.0397, 0.0299, 0.0653, 0.0463, 0.0956, 0.0943, 0.0332], + device='cuda:1'), in_proj_covar=tensor([0.0458, 0.0399, 0.0351, 0.0449, 0.0382, 0.0539, 0.0393, 0.0426], + device='cuda:1'), out_proj_covar=tensor([1.2231e-04, 1.0434e-04, 9.2115e-05, 1.1794e-04, 1.0032e-04, 1.5167e-04, + 1.0568e-04, 1.1243e-04], device='cuda:1') +2023-02-07 06:27:01,112 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186492.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:01,711 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186493.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:03,579 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.350e+02 3.005e+02 3.846e+02 7.955e+02, threshold=6.011e+02, percent-clipped=1.0 +2023-02-07 06:27:12,588 INFO [train.py:901] (1/4) Epoch 24, batch 600, loss[loss=0.2432, simple_loss=0.3134, pruned_loss=0.08652, over 8032.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2886, pruned_loss=0.06188, over 1542699.89 frames. ], batch size: 22, lr: 3.17e-03, grad_scale: 16.0 +2023-02-07 06:27:19,620 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:21,528 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186520.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:23,334 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.91 vs. limit=5.0 +2023-02-07 06:27:26,202 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 06:27:29,790 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186532.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 06:27:31,744 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186535.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:40,305 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7301, 1.3971, 2.8387, 1.4059, 2.1846, 3.0338, 3.1378, 2.5880], + device='cuda:1'), covar=tensor([0.1134, 0.1738, 0.0385, 0.2104, 0.0974, 0.0295, 0.0749, 0.0594], + device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0321, 0.0286, 0.0314, 0.0312, 0.0268, 0.0424, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:27:46,480 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186557.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:46,550 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186557.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 06:27:47,027 INFO [train.py:901] (1/4) Epoch 24, batch 650, loss[loss=0.2038, simple_loss=0.293, pruned_loss=0.05733, over 8524.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2876, pruned_loss=0.06159, over 1557437.52 frames. ], batch size: 39, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:27:52,102 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9108, 1.7060, 3.5669, 1.5993, 2.4876, 3.9185, 3.9817, 3.3591], + device='cuda:1'), covar=tensor([0.1270, 0.1691, 0.0299, 0.1996, 0.0989, 0.0209, 0.0505, 0.0516], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0321, 0.0285, 0.0314, 0.0311, 0.0268, 0.0423, 0.0302], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:28:01,234 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186577.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:07,495 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186585.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:13,142 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3225, 1.7941, 3.3775, 1.4486, 2.4281, 3.7346, 3.8406, 3.2023], + device='cuda:1'), covar=tensor([0.0979, 0.1695, 0.0407, 0.2237, 0.1215, 0.0241, 0.0534, 0.0557], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0320, 0.0284, 0.0312, 0.0310, 0.0267, 0.0422, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:28:15,661 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 2.377e+02 2.753e+02 3.513e+02 8.271e+02, threshold=5.505e+02, percent-clipped=2.0 +2023-02-07 06:28:20,753 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186604.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:23,404 INFO [train.py:901] (1/4) Epoch 24, batch 700, loss[loss=0.2047, simple_loss=0.2888, pruned_loss=0.0603, over 8336.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2877, pruned_loss=0.06114, over 1575035.22 frames. ], batch size: 25, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:28:24,190 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2115, 4.1733, 3.8090, 1.9490, 3.7234, 3.7549, 3.7227, 3.6234], + device='cuda:1'), covar=tensor([0.0747, 0.0560, 0.0953, 0.4448, 0.0866, 0.0972, 0.1247, 0.0778], + device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0446, 0.0432, 0.0543, 0.0430, 0.0448, 0.0427, 0.0389], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:28:33,222 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186621.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:43,730 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186635.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:59,666 INFO [train.py:901] (1/4) Epoch 24, batch 750, loss[loss=0.203, simple_loss=0.2872, pruned_loss=0.05935, over 7819.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2867, pruned_loss=0.06084, over 1582556.48 frames. ], batch size: 20, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:29:00,494 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5975, 1.9030, 2.0335, 1.1351, 2.1499, 1.3492, 0.7027, 1.7871], + device='cuda:1'), covar=tensor([0.0766, 0.0416, 0.0314, 0.0740, 0.0463, 0.1098, 0.1039, 0.0420], + device='cuda:1'), in_proj_covar=tensor([0.0457, 0.0397, 0.0350, 0.0447, 0.0379, 0.0536, 0.0391, 0.0425], + device='cuda:1'), out_proj_covar=tensor([1.2187e-04, 1.0376e-04, 9.1745e-05, 1.1728e-04, 9.9550e-05, 1.5077e-04, + 1.0520e-04, 1.1211e-04], device='cuda:1') +2023-02-07 06:29:11,910 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 06:29:16,223 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1752, 1.6548, 4.6794, 1.8964, 3.7272, 3.7608, 4.1681, 4.1771], + device='cuda:1'), covar=tensor([0.1376, 0.6218, 0.0931, 0.5071, 0.2129, 0.1573, 0.1023, 0.0964], + device='cuda:1'), in_proj_covar=tensor([0.0645, 0.0653, 0.0707, 0.0637, 0.0721, 0.0620, 0.0613, 0.0689], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:29:21,537 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 06:29:27,090 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.532e+02 3.077e+02 4.008e+02 9.294e+02, threshold=6.153e+02, percent-clipped=8.0 +2023-02-07 06:29:31,913 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 06:29:35,745 INFO [train.py:901] (1/4) Epoch 24, batch 800, loss[loss=0.2095, simple_loss=0.3068, pruned_loss=0.05609, over 8336.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2864, pruned_loss=0.06029, over 1593749.28 frames. ], batch size: 26, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:29:48,932 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186727.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:29:56,176 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186736.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:30:11,904 INFO [train.py:901] (1/4) Epoch 24, batch 850, loss[loss=0.2107, simple_loss=0.2951, pruned_loss=0.06318, over 7808.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2864, pruned_loss=0.06003, over 1602509.93 frames. ], batch size: 20, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:30:29,493 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186783.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:30:39,064 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.330e+02 2.764e+02 3.350e+02 7.186e+02, threshold=5.528e+02, percent-clipped=2.0 +2023-02-07 06:30:47,640 INFO [train.py:901] (1/4) Epoch 24, batch 900, loss[loss=0.1862, simple_loss=0.2754, pruned_loss=0.0485, over 8451.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2864, pruned_loss=0.05995, over 1608600.99 frames. ], batch size: 27, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:31:05,961 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186833.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:08,582 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186837.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:15,813 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6779, 2.4804, 3.2731, 2.5411, 3.2792, 2.7249, 2.5848, 2.0354], + device='cuda:1'), covar=tensor([0.5037, 0.4628, 0.1813, 0.4010, 0.2520, 0.2844, 0.1709, 0.5303], + device='cuda:1'), in_proj_covar=tensor([0.0948, 0.0992, 0.0812, 0.0958, 0.0998, 0.0901, 0.0756, 0.0830], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 06:31:24,266 INFO [train.py:901] (1/4) Epoch 24, batch 950, loss[loss=0.2508, simple_loss=0.3207, pruned_loss=0.09051, over 7269.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2857, pruned_loss=0.05993, over 1604127.00 frames. ], batch size: 73, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:31:24,446 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186858.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:24,508 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186858.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:30,139 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2437, 1.5897, 1.7675, 1.4674, 1.0728, 1.5903, 1.8647, 1.9599], + device='cuda:1'), covar=tensor([0.0495, 0.1215, 0.1688, 0.1418, 0.0614, 0.1455, 0.0674, 0.0574], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0159, 0.0100, 0.0162, 0.0111, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 06:31:39,987 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186879.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:43,467 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 06:31:48,341 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186891.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:50,396 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3348, 2.1360, 1.7052, 1.9953, 1.7304, 1.4729, 1.7370, 1.7168], + device='cuda:1'), covar=tensor([0.1179, 0.0383, 0.1186, 0.0499, 0.0733, 0.1506, 0.0902, 0.0776], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0234, 0.0336, 0.0310, 0.0301, 0.0341, 0.0346, 0.0319], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 06:31:52,263 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.527e+02 2.324e+02 2.850e+02 3.567e+02 7.043e+02, threshold=5.700e+02, percent-clipped=2.0 +2023-02-07 06:31:53,138 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186898.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:55,084 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186901.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:59,853 INFO [train.py:901] (1/4) Epoch 24, batch 1000, loss[loss=0.2115, simple_loss=0.3002, pruned_loss=0.0614, over 8350.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2854, pruned_loss=0.06007, over 1604905.86 frames. ], batch size: 24, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:32:05,643 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186916.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:15,390 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186929.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:15,466 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186929.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:19,246 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5819, 2.6017, 1.7864, 2.3258, 2.1592, 1.5789, 2.1467, 2.2453], + device='cuda:1'), covar=tensor([0.1761, 0.0500, 0.1409, 0.0769, 0.0858, 0.1653, 0.1153, 0.1239], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0235, 0.0336, 0.0310, 0.0301, 0.0341, 0.0347, 0.0319], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 06:32:20,501 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 06:32:29,428 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186948.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:32,162 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186952.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:33,326 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 06:32:36,111 INFO [train.py:901] (1/4) Epoch 24, batch 1050, loss[loss=0.1819, simple_loss=0.2589, pruned_loss=0.0525, over 7805.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2852, pruned_loss=0.06012, over 1607049.31 frames. ], batch size: 20, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:33:00,036 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7837, 1.4876, 3.9751, 1.5258, 3.5144, 3.2896, 3.6111, 3.4901], + device='cuda:1'), covar=tensor([0.0717, 0.4510, 0.0742, 0.4123, 0.1310, 0.1151, 0.0672, 0.0759], + device='cuda:1'), in_proj_covar=tensor([0.0642, 0.0649, 0.0703, 0.0634, 0.0718, 0.0618, 0.0610, 0.0686], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:33:01,540 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186992.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:02,866 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186994.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:04,734 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.475e+02 2.949e+02 3.829e+02 9.793e+02, threshold=5.897e+02, percent-clipped=8.0 +2023-02-07 06:33:12,436 INFO [train.py:901] (1/4) Epoch 24, batch 1100, loss[loss=0.1785, simple_loss=0.2656, pruned_loss=0.04566, over 7555.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2861, pruned_loss=0.06051, over 1613073.41 frames. ], batch size: 18, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:33:12,699 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5579, 2.2559, 4.0425, 1.4382, 2.9064, 2.0432, 1.8306, 2.8512], + device='cuda:1'), covar=tensor([0.2087, 0.2779, 0.0828, 0.4962, 0.1913, 0.3443, 0.2493, 0.2553], + device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0615, 0.0556, 0.0650, 0.0653, 0.0600, 0.0544, 0.0634], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:33:18,326 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187016.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:19,043 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187017.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:37,509 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187043.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:38,291 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187044.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:45,711 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 06:33:47,123 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187056.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:48,353 INFO [train.py:901] (1/4) Epoch 24, batch 1150, loss[loss=0.1961, simple_loss=0.288, pruned_loss=0.05211, over 8493.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2871, pruned_loss=0.06098, over 1613117.44 frames. ], batch size: 28, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:33:49,926 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9199, 1.5305, 3.3764, 1.4808, 2.4006, 3.6290, 3.7449, 3.1037], + device='cuda:1'), covar=tensor([0.1179, 0.1771, 0.0307, 0.1964, 0.0967, 0.0231, 0.0569, 0.0526], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0324, 0.0288, 0.0317, 0.0315, 0.0270, 0.0427, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:33:52,032 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187063.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:57,505 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187071.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:34:02,538 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0459, 1.6086, 1.4163, 1.5556, 1.3269, 1.3107, 1.3242, 1.3064], + device='cuda:1'), covar=tensor([0.1052, 0.0492, 0.1266, 0.0545, 0.0737, 0.1438, 0.0872, 0.0801], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0234, 0.0336, 0.0309, 0.0301, 0.0341, 0.0346, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 06:34:16,148 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.205e+02 2.742e+02 3.279e+02 6.267e+02, threshold=5.485e+02, percent-clipped=2.0 +2023-02-07 06:34:24,623 INFO [train.py:901] (1/4) Epoch 24, batch 1200, loss[loss=0.2356, simple_loss=0.313, pruned_loss=0.07904, over 7976.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2868, pruned_loss=0.06062, over 1617058.43 frames. ], batch size: 21, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:34:57,398 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187154.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:35:00,007 INFO [train.py:901] (1/4) Epoch 24, batch 1250, loss[loss=0.1936, simple_loss=0.2813, pruned_loss=0.05301, over 8088.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2864, pruned_loss=0.06039, over 1618073.89 frames. ], batch size: 21, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:35:15,148 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187179.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:35:19,765 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187186.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:35:27,953 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.417e+02 2.916e+02 3.659e+02 9.833e+02, threshold=5.832e+02, percent-clipped=6.0 +2023-02-07 06:35:30,894 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187202.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:35:35,736 INFO [train.py:901] (1/4) Epoch 24, batch 1300, loss[loss=0.2122, simple_loss=0.3133, pruned_loss=0.05559, over 8196.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2858, pruned_loss=0.05986, over 1620177.96 frames. ], batch size: 23, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:35:35,964 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187208.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:35:51,426 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.94 vs. limit=5.0 +2023-02-07 06:35:53,889 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187233.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:05,770 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187250.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:10,944 INFO [train.py:901] (1/4) Epoch 24, batch 1350, loss[loss=0.2144, simple_loss=0.2953, pruned_loss=0.06676, over 8508.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2853, pruned_loss=0.05945, over 1618899.58 frames. ], batch size: 28, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:36:20,761 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187272.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:21,318 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187273.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:22,816 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187275.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:39,304 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187297.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:39,750 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.841e+02 2.391e+02 3.088e+02 3.702e+02 1.176e+03, threshold=6.175e+02, percent-clipped=8.0 +2023-02-07 06:36:41,384 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187300.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:46,710 INFO [train.py:901] (1/4) Epoch 24, batch 1400, loss[loss=0.1937, simple_loss=0.2675, pruned_loss=0.05993, over 7693.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2854, pruned_loss=0.05979, over 1620245.70 frames. ], batch size: 18, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:36:52,960 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187317.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:54,421 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187319.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:59,292 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187325.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:03,362 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3843, 1.6197, 2.0963, 1.3456, 1.4975, 1.6665, 1.4953, 1.4287], + device='cuda:1'), covar=tensor([0.2010, 0.2409, 0.0973, 0.4487, 0.1923, 0.3424, 0.2420, 0.2215], + device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0617, 0.0558, 0.0652, 0.0653, 0.0600, 0.0545, 0.0636], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:37:05,353 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5948, 2.4761, 1.7855, 2.2578, 2.0325, 1.5266, 2.0604, 2.0741], + device='cuda:1'), covar=tensor([0.1537, 0.0421, 0.1232, 0.0624, 0.0785, 0.1563, 0.1062, 0.0967], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0233, 0.0335, 0.0308, 0.0300, 0.0338, 0.0346, 0.0316], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 06:37:12,877 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187344.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:16,708 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-07 06:37:21,717 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 06:37:22,426 INFO [train.py:901] (1/4) Epoch 24, batch 1450, loss[loss=0.2103, simple_loss=0.3023, pruned_loss=0.05909, over 8514.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2864, pruned_loss=0.06089, over 1620532.02 frames. ], batch size: 28, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:37:42,314 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187387.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:43,074 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187388.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:44,410 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187390.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:49,534 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.482e+02 2.870e+02 4.012e+02 8.494e+02, threshold=5.740e+02, percent-clipped=8.0 +2023-02-07 06:37:51,823 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187400.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:56,995 INFO [train.py:901] (1/4) Epoch 24, batch 1500, loss[loss=0.1731, simple_loss=0.2562, pruned_loss=0.04499, over 8247.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.286, pruned_loss=0.06075, over 1618739.97 frames. ], batch size: 22, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:38:22,159 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187442.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:38:33,246 INFO [train.py:901] (1/4) Epoch 24, batch 1550, loss[loss=0.194, simple_loss=0.286, pruned_loss=0.05103, over 8515.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2865, pruned_loss=0.06076, over 1620008.29 frames. ], batch size: 28, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:38:39,573 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187467.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:38:59,897 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.355e+02 2.764e+02 3.622e+02 7.454e+02, threshold=5.529e+02, percent-clipped=4.0 +2023-02-07 06:39:02,833 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187502.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:39:06,539 INFO [train.py:901] (1/4) Epoch 24, batch 1600, loss[loss=0.1736, simple_loss=0.2535, pruned_loss=0.04684, over 7715.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2872, pruned_loss=0.06059, over 1626001.38 frames. ], batch size: 18, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:39:11,516 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187515.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:39:22,051 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187530.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:39:27,624 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187537.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:39:30,381 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187541.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:39:41,438 INFO [train.py:901] (1/4) Epoch 24, batch 1650, loss[loss=0.2043, simple_loss=0.2925, pruned_loss=0.05804, over 8605.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2867, pruned_loss=0.06026, over 1626277.44 frames. ], batch size: 34, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:39:52,523 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187573.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:40:09,686 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.451e+02 2.921e+02 3.516e+02 7.853e+02, threshold=5.842e+02, percent-clipped=7.0 +2023-02-07 06:40:09,874 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187598.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:40:10,449 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9713, 1.3647, 4.4897, 2.1381, 2.4842, 5.0956, 5.1985, 4.4473], + device='cuda:1'), covar=tensor([0.1279, 0.1900, 0.0240, 0.1861, 0.1127, 0.0171, 0.0346, 0.0513], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0319, 0.0283, 0.0312, 0.0311, 0.0267, 0.0422, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:40:16,433 INFO [train.py:901] (1/4) Epoch 24, batch 1700, loss[loss=0.197, simple_loss=0.2829, pruned_loss=0.05555, over 8357.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2865, pruned_loss=0.06051, over 1623379.89 frames. ], batch size: 24, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:40:40,843 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187644.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:40:50,898 INFO [train.py:901] (1/4) Epoch 24, batch 1750, loss[loss=0.2273, simple_loss=0.307, pruned_loss=0.07378, over 8443.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.287, pruned_loss=0.06105, over 1619705.38 frames. ], batch size: 49, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:40:58,584 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187669.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:41:18,565 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.502e+02 3.000e+02 3.757e+02 9.885e+02, threshold=5.999e+02, percent-clipped=2.0 +2023-02-07 06:41:25,257 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 06:41:26,166 INFO [train.py:901] (1/4) Epoch 24, batch 1800, loss[loss=0.1809, simple_loss=0.2554, pruned_loss=0.05313, over 7521.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2863, pruned_loss=0.06063, over 1621016.04 frames. ], batch size: 18, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:41:43,267 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187734.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:41:59,216 INFO [train.py:901] (1/4) Epoch 24, batch 1850, loss[loss=0.2222, simple_loss=0.2868, pruned_loss=0.07878, over 7517.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2853, pruned_loss=0.06011, over 1618933.64 frames. ], batch size: 18, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:41:59,466 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187758.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:42:09,326 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187771.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:42:14,901 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6588, 1.4936, 2.8698, 1.4547, 2.2493, 3.1000, 3.2173, 2.6991], + device='cuda:1'), covar=tensor([0.1157, 0.1534, 0.0377, 0.2034, 0.0889, 0.0297, 0.0612, 0.0540], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0319, 0.0284, 0.0312, 0.0312, 0.0267, 0.0423, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:42:18,477 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187783.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:42:27,424 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187796.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:42:28,603 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 2.492e+02 2.912e+02 4.002e+02 8.326e+02, threshold=5.824e+02, percent-clipped=6.0 +2023-02-07 06:42:36,389 INFO [train.py:901] (1/4) Epoch 24, batch 1900, loss[loss=0.1979, simple_loss=0.2733, pruned_loss=0.06125, over 7811.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2851, pruned_loss=0.05991, over 1618968.41 frames. ], batch size: 20, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:42:42,667 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.5705, 5.5813, 4.9133, 2.4566, 4.9621, 5.3877, 5.1912, 5.2402], + device='cuda:1'), covar=tensor([0.0580, 0.0413, 0.0923, 0.4780, 0.0822, 0.0964, 0.1067, 0.0649], + device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0446, 0.0435, 0.0543, 0.0435, 0.0448, 0.0428, 0.0391], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:42:49,918 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 +2023-02-07 06:43:02,032 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 06:43:05,644 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 06:43:05,819 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187849.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:43:11,794 INFO [train.py:901] (1/4) Epoch 24, batch 1950, loss[loss=0.1608, simple_loss=0.2496, pruned_loss=0.03595, over 8035.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2854, pruned_loss=0.06002, over 1620711.10 frames. ], batch size: 22, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:43:18,398 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 06:43:22,598 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187874.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:43:27,989 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187881.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:43:30,734 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187885.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:43:39,199 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 2.375e+02 2.745e+02 3.412e+02 6.105e+02, threshold=5.491e+02, percent-clipped=1.0 +2023-02-07 06:43:39,235 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 06:43:42,665 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3984, 1.7349, 1.8149, 1.0751, 1.8579, 1.3328, 0.4300, 1.6412], + device='cuda:1'), covar=tensor([0.0768, 0.0449, 0.0422, 0.0708, 0.0565, 0.1179, 0.1082, 0.0393], + device='cuda:1'), in_proj_covar=tensor([0.0458, 0.0401, 0.0355, 0.0451, 0.0385, 0.0541, 0.0394, 0.0426], + device='cuda:1'), out_proj_covar=tensor([1.2231e-04, 1.0479e-04, 9.3321e-05, 1.1851e-04, 1.0101e-04, 1.5216e-04, + 1.0605e-04, 1.1231e-04], device='cuda:1') +2023-02-07 06:43:46,287 INFO [train.py:901] (1/4) Epoch 24, batch 2000, loss[loss=0.2181, simple_loss=0.3042, pruned_loss=0.06607, over 8458.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2846, pruned_loss=0.05966, over 1619843.25 frames. ], batch size: 25, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:44:21,537 INFO [train.py:901] (1/4) Epoch 24, batch 2050, loss[loss=0.1955, simple_loss=0.2867, pruned_loss=0.05215, over 8473.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2847, pruned_loss=0.05924, over 1619117.00 frames. ], batch size: 29, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:44:23,781 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7934, 2.0104, 2.2062, 1.3173, 2.2820, 1.6658, 0.7651, 1.8995], + device='cuda:1'), covar=tensor([0.0612, 0.0379, 0.0296, 0.0667, 0.0456, 0.0940, 0.0931, 0.0365], + device='cuda:1'), in_proj_covar=tensor([0.0458, 0.0400, 0.0354, 0.0450, 0.0384, 0.0539, 0.0394, 0.0425], + device='cuda:1'), out_proj_covar=tensor([1.2219e-04, 1.0450e-04, 9.3072e-05, 1.1817e-04, 1.0087e-04, 1.5174e-04, + 1.0582e-04, 1.1198e-04], device='cuda:1') +2023-02-07 06:44:42,289 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187989.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:44:46,914 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187996.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:44:48,823 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.387e+02 2.958e+02 3.531e+02 6.524e+02, threshold=5.915e+02, percent-clipped=3.0 +2023-02-07 06:44:51,434 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188000.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:44:56,709 INFO [train.py:901] (1/4) Epoch 24, batch 2100, loss[loss=0.2311, simple_loss=0.3025, pruned_loss=0.0798, over 8506.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2856, pruned_loss=0.05977, over 1619082.05 frames. ], batch size: 28, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:44:58,661 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-02-07 06:45:31,688 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8005, 2.1299, 3.5498, 1.8148, 1.6812, 3.5201, 0.6609, 2.1548], + device='cuda:1'), covar=tensor([0.1351, 0.1357, 0.0255, 0.1604, 0.2704, 0.0268, 0.2296, 0.1156], + device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0200, 0.0132, 0.0222, 0.0273, 0.0137, 0.0172, 0.0195], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 06:45:32,168 INFO [train.py:901] (1/4) Epoch 24, batch 2150, loss[loss=0.2138, simple_loss=0.2989, pruned_loss=0.06433, over 8426.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.286, pruned_loss=0.06028, over 1615209.57 frames. ], batch size: 27, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:45:47,939 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2855, 1.9061, 2.4480, 2.1335, 2.4183, 2.3318, 2.1518, 1.2625], + device='cuda:1'), covar=tensor([0.5737, 0.5138, 0.2072, 0.3526, 0.2393, 0.3033, 0.1873, 0.5267], + device='cuda:1'), in_proj_covar=tensor([0.0951, 0.1000, 0.0818, 0.0965, 0.1002, 0.0908, 0.0761, 0.0835], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 06:45:58,768 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.464e+02 3.048e+02 3.692e+02 7.821e+02, threshold=6.095e+02, percent-clipped=5.0 +2023-02-07 06:46:03,898 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188105.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:46:05,700 INFO [train.py:901] (1/4) Epoch 24, batch 2200, loss[loss=0.2193, simple_loss=0.3011, pruned_loss=0.06876, over 8507.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2858, pruned_loss=0.06001, over 1615947.21 frames. ], batch size: 26, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:46:21,943 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188130.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:46:41,366 INFO [train.py:901] (1/4) Epoch 24, batch 2250, loss[loss=0.2149, simple_loss=0.2925, pruned_loss=0.06867, over 8640.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2852, pruned_loss=0.05923, over 1617320.59 frames. ], batch size: 34, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:47:09,411 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.349e+02 3.047e+02 3.898e+02 9.680e+02, threshold=6.095e+02, percent-clipped=4.0 +2023-02-07 06:47:16,302 INFO [train.py:901] (1/4) Epoch 24, batch 2300, loss[loss=0.177, simple_loss=0.2571, pruned_loss=0.04838, over 7805.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2854, pruned_loss=0.05929, over 1614865.06 frames. ], batch size: 20, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:47:42,353 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188245.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:47:47,128 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188252.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:47:50,574 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188256.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:47:51,703 INFO [train.py:901] (1/4) Epoch 24, batch 2350, loss[loss=0.1916, simple_loss=0.2608, pruned_loss=0.0612, over 7236.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2849, pruned_loss=0.05921, over 1610361.34 frames. ], batch size: 16, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:48:00,161 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188270.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:48:05,481 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188277.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:48:08,165 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188281.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:48:20,112 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.514e+02 3.085e+02 3.939e+02 8.316e+02, threshold=6.171e+02, percent-clipped=4.0 +2023-02-07 06:48:22,397 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188301.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:48:27,100 INFO [train.py:901] (1/4) Epoch 24, batch 2400, loss[loss=0.223, simple_loss=0.3049, pruned_loss=0.07049, over 8325.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2858, pruned_loss=0.06018, over 1608789.33 frames. ], batch size: 25, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:48:30,048 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8144, 1.6166, 3.1107, 1.6301, 2.2474, 3.3417, 3.5078, 2.9287], + device='cuda:1'), covar=tensor([0.1235, 0.1634, 0.0364, 0.1951, 0.1085, 0.0275, 0.0522, 0.0514], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0323, 0.0285, 0.0315, 0.0314, 0.0270, 0.0425, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 06:49:00,630 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.87 vs. limit=5.0 +2023-02-07 06:49:02,243 INFO [train.py:901] (1/4) Epoch 24, batch 2450, loss[loss=0.1931, simple_loss=0.2839, pruned_loss=0.05121, over 8291.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2864, pruned_loss=0.06039, over 1612383.82 frames. ], batch size: 23, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:49:30,918 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 2.263e+02 2.982e+02 3.612e+02 7.179e+02, threshold=5.965e+02, percent-clipped=1.0 +2023-02-07 06:49:38,102 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-07 06:49:38,406 INFO [train.py:901] (1/4) Epoch 24, batch 2500, loss[loss=0.182, simple_loss=0.2719, pruned_loss=0.04607, over 8588.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2859, pruned_loss=0.06039, over 1613627.63 frames. ], batch size: 49, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:50:11,883 INFO [train.py:901] (1/4) Epoch 24, batch 2550, loss[loss=0.1629, simple_loss=0.2609, pruned_loss=0.03239, over 5536.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2866, pruned_loss=0.06079, over 1610729.40 frames. ], batch size: 12, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:50:40,453 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.347e+02 2.538e+02 2.905e+02 3.766e+02 9.788e+02, threshold=5.809e+02, percent-clipped=4.0 +2023-02-07 06:50:41,315 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188499.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:50:47,916 INFO [train.py:901] (1/4) Epoch 24, batch 2600, loss[loss=0.2087, simple_loss=0.3011, pruned_loss=0.05819, over 8301.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2858, pruned_loss=0.06046, over 1612328.26 frames. ], batch size: 23, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:51:21,918 INFO [train.py:901] (1/4) Epoch 24, batch 2650, loss[loss=0.1991, simple_loss=0.2837, pruned_loss=0.05729, over 8040.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2855, pruned_loss=0.06077, over 1611111.68 frames. ], batch size: 22, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:51:31,890 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-02-07 06:51:41,776 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.85 vs. limit=5.0 +2023-02-07 06:51:48,605 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 2.338e+02 2.924e+02 3.924e+02 7.774e+02, threshold=5.847e+02, percent-clipped=4.0 +2023-02-07 06:51:55,411 INFO [train.py:901] (1/4) Epoch 24, batch 2700, loss[loss=0.161, simple_loss=0.2565, pruned_loss=0.03281, over 7807.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.286, pruned_loss=0.06069, over 1611820.12 frames. ], batch size: 20, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:52:02,966 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7971, 5.9724, 5.1625, 2.6812, 5.2820, 5.6633, 5.3406, 5.4744], + device='cuda:1'), covar=tensor([0.0571, 0.0368, 0.0869, 0.4082, 0.0709, 0.0711, 0.1062, 0.0446], + device='cuda:1'), in_proj_covar=tensor([0.0521, 0.0439, 0.0426, 0.0535, 0.0426, 0.0441, 0.0421, 0.0386], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:52:21,731 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=188645.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:52:31,277 INFO [train.py:901] (1/4) Epoch 24, batch 2750, loss[loss=0.2309, simple_loss=0.3073, pruned_loss=0.07728, over 8539.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2852, pruned_loss=0.06061, over 1611758.74 frames. ], batch size: 39, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:52:57,784 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 2.375e+02 3.087e+02 4.139e+02 1.460e+03, threshold=6.174e+02, percent-clipped=4.0 +2023-02-07 06:53:05,388 INFO [train.py:901] (1/4) Epoch 24, batch 2800, loss[loss=0.2091, simple_loss=0.3046, pruned_loss=0.05677, over 8784.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2853, pruned_loss=0.06078, over 1612802.62 frames. ], batch size: 30, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:53:23,062 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9053, 1.6673, 2.0513, 1.7986, 2.0083, 1.9813, 1.8109, 0.8297], + device='cuda:1'), covar=tensor([0.5897, 0.4849, 0.2121, 0.3781, 0.2483, 0.3303, 0.2059, 0.5205], + device='cuda:1'), in_proj_covar=tensor([0.0947, 0.0992, 0.0814, 0.0956, 0.0995, 0.0905, 0.0756, 0.0832], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 06:53:40,615 INFO [train.py:901] (1/4) Epoch 24, batch 2850, loss[loss=0.1738, simple_loss=0.251, pruned_loss=0.0483, over 7203.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2862, pruned_loss=0.06105, over 1612475.08 frames. ], batch size: 16, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:53:42,193 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188760.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:54:07,931 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.346e+02 3.064e+02 3.754e+02 6.997e+02, threshold=6.129e+02, percent-clipped=3.0 +2023-02-07 06:54:14,861 INFO [train.py:901] (1/4) Epoch 24, batch 2900, loss[loss=0.1913, simple_loss=0.2764, pruned_loss=0.05308, over 7982.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2865, pruned_loss=0.06151, over 1611886.26 frames. ], batch size: 21, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:54:39,482 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=188843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:54:48,810 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4468, 1.5639, 1.4238, 1.8082, 0.7363, 1.3373, 1.3281, 1.5054], + device='cuda:1'), covar=tensor([0.0860, 0.0726, 0.0978, 0.0505, 0.1105, 0.1297, 0.0703, 0.0760], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0197, 0.0242, 0.0214, 0.0205, 0.0246, 0.0249, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 06:54:51,404 INFO [train.py:901] (1/4) Epoch 24, batch 2950, loss[loss=0.2199, simple_loss=0.3073, pruned_loss=0.06623, over 8261.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2858, pruned_loss=0.0607, over 1611746.02 frames. ], batch size: 24, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:54:51,417 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 06:55:19,115 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.217e+02 2.685e+02 3.700e+02 9.567e+02, threshold=5.370e+02, percent-clipped=4.0 +2023-02-07 06:55:25,890 INFO [train.py:901] (1/4) Epoch 24, batch 3000, loss[loss=0.1786, simple_loss=0.2602, pruned_loss=0.04847, over 7974.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2856, pruned_loss=0.06082, over 1608214.64 frames. ], batch size: 21, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:55:25,890 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 06:55:39,550 INFO [train.py:935] (1/4) Epoch 24, validation: loss=0.1724, simple_loss=0.2726, pruned_loss=0.03604, over 944034.00 frames. +2023-02-07 06:55:39,551 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 06:56:05,990 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188947.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:56:07,365 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188949.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:56:13,963 INFO [train.py:901] (1/4) Epoch 24, batch 3050, loss[loss=0.2306, simple_loss=0.3196, pruned_loss=0.07078, over 8504.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2869, pruned_loss=0.06134, over 1609830.08 frames. ], batch size: 26, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:56:14,162 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188958.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:56:41,563 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.422e+02 3.010e+02 3.817e+02 9.746e+02, threshold=6.020e+02, percent-clipped=4.0 +2023-02-07 06:56:49,123 INFO [train.py:901] (1/4) Epoch 24, batch 3100, loss[loss=0.2368, simple_loss=0.332, pruned_loss=0.07083, over 8267.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.287, pruned_loss=0.06103, over 1611311.58 frames. ], batch size: 24, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:56:54,800 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189016.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:56:58,627 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4968, 4.5153, 4.0048, 2.1686, 3.9794, 4.0356, 3.9988, 3.8677], + device='cuda:1'), covar=tensor([0.0674, 0.0473, 0.0934, 0.4529, 0.0889, 0.0930, 0.1197, 0.0613], + device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0442, 0.0430, 0.0542, 0.0429, 0.0444, 0.0424, 0.0389], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 06:57:12,105 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189041.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:57:16,988 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189048.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:57:19,123 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6889, 1.6400, 2.4576, 1.5805, 1.3048, 2.3496, 0.6934, 1.5402], + device='cuda:1'), covar=tensor([0.1419, 0.1185, 0.0304, 0.1217, 0.2523, 0.0382, 0.1997, 0.1221], + device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0202, 0.0132, 0.0223, 0.0274, 0.0139, 0.0173, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 06:57:23,529 INFO [train.py:901] (1/4) Epoch 24, batch 3150, loss[loss=0.2168, simple_loss=0.3073, pruned_loss=0.06313, over 8468.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2871, pruned_loss=0.06121, over 1608096.96 frames. ], batch size: 25, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:57:26,636 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.74 vs. limit=5.0 +2023-02-07 06:57:40,665 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189082.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:57:50,853 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189097.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:57:51,369 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.687e+02 2.394e+02 2.917e+02 3.565e+02 6.979e+02, threshold=5.834e+02, percent-clipped=3.0 +2023-02-07 06:57:59,734 INFO [train.py:901] (1/4) Epoch 24, batch 3200, loss[loss=0.193, simple_loss=0.2702, pruned_loss=0.05793, over 7656.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2876, pruned_loss=0.06073, over 1616914.65 frames. ], batch size: 19, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:58:33,951 INFO [train.py:901] (1/4) Epoch 24, batch 3250, loss[loss=0.2098, simple_loss=0.2878, pruned_loss=0.06597, over 7930.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2896, pruned_loss=0.06212, over 1620254.12 frames. ], batch size: 20, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 06:59:01,531 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.438e+02 3.003e+02 3.759e+02 6.490e+02, threshold=6.005e+02, percent-clipped=4.0 +2023-02-07 06:59:08,523 INFO [train.py:901] (1/4) Epoch 24, batch 3300, loss[loss=0.2395, simple_loss=0.3067, pruned_loss=0.08613, over 7038.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2881, pruned_loss=0.0615, over 1614961.83 frames. ], batch size: 73, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 06:59:12,860 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:59:30,915 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:59:39,008 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6749, 2.6864, 1.8379, 2.3593, 2.2632, 1.6136, 2.1428, 2.2483], + device='cuda:1'), covar=tensor([0.1559, 0.0379, 0.1276, 0.0679, 0.0706, 0.1613, 0.1107, 0.1001], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0235, 0.0334, 0.0310, 0.0299, 0.0341, 0.0345, 0.0317], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 06:59:44,159 INFO [train.py:901] (1/4) Epoch 24, batch 3350, loss[loss=0.2417, simple_loss=0.3128, pruned_loss=0.08529, over 8845.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2883, pruned_loss=0.06136, over 1619438.86 frames. ], batch size: 51, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 06:59:49,402 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189266.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:59:54,641 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8681, 1.9389, 3.2644, 2.3720, 2.8635, 1.8646, 1.6357, 1.7672], + device='cuda:1'), covar=tensor([0.8066, 0.6755, 0.2013, 0.4660, 0.3561, 0.5125, 0.3430, 0.5966], + device='cuda:1'), in_proj_covar=tensor([0.0945, 0.0991, 0.0811, 0.0955, 0.0994, 0.0902, 0.0754, 0.0828], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 07:00:05,778 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189291.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:00:07,042 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189293.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:00:10,334 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.367e+02 2.967e+02 3.575e+02 9.298e+02, threshold=5.934e+02, percent-clipped=5.0 +2023-02-07 07:00:17,739 INFO [train.py:901] (1/4) Epoch 24, batch 3400, loss[loss=0.2421, simple_loss=0.3312, pruned_loss=0.07653, over 8330.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.289, pruned_loss=0.06189, over 1620462.03 frames. ], batch size: 25, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:00:33,355 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5940, 1.3326, 2.9106, 1.1511, 2.3041, 3.1360, 3.4445, 2.3329], + device='cuda:1'), covar=tensor([0.1681, 0.2196, 0.0550, 0.3037, 0.1192, 0.0468, 0.0676, 0.1101], + device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0324, 0.0286, 0.0317, 0.0316, 0.0271, 0.0429, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 07:00:34,013 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0078, 1.6606, 1.7582, 1.5242, 0.8616, 1.6169, 1.7270, 1.6950], + device='cuda:1'), covar=tensor([0.0502, 0.1174, 0.1587, 0.1309, 0.0571, 0.1348, 0.0664, 0.0571], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0160, 0.0100, 0.0163, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 07:00:52,447 INFO [train.py:901] (1/4) Epoch 24, batch 3450, loss[loss=0.199, simple_loss=0.2747, pruned_loss=0.06168, over 8243.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2864, pruned_loss=0.06107, over 1615548.62 frames. ], batch size: 22, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:00:57,504 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189365.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:16,612 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189392.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:20,611 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.399e+02 2.884e+02 3.624e+02 7.571e+02, threshold=5.767e+02, percent-clipped=3.0 +2023-02-07 07:01:26,256 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189406.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:27,409 INFO [train.py:901] (1/4) Epoch 24, batch 3500, loss[loss=0.2213, simple_loss=0.3061, pruned_loss=0.06826, over 8752.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2877, pruned_loss=0.06137, over 1617527.28 frames. ], batch size: 30, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:01:27,625 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189408.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:27,666 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7154, 1.9952, 2.0826, 1.4852, 2.2543, 1.5725, 0.7848, 1.9174], + device='cuda:1'), covar=tensor([0.0710, 0.0373, 0.0352, 0.0597, 0.0515, 0.0872, 0.0934, 0.0392], + device='cuda:1'), in_proj_covar=tensor([0.0461, 0.0401, 0.0356, 0.0454, 0.0387, 0.0542, 0.0398, 0.0431], + device='cuda:1'), out_proj_covar=tensor([1.2292e-04, 1.0469e-04, 9.3477e-05, 1.1928e-04, 1.0174e-04, 1.5223e-04, + 1.0689e-04, 1.1361e-04], device='cuda:1') +2023-02-07 07:01:40,504 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 07:01:40,597 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189426.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:50,793 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189441.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:56,339 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 +2023-02-07 07:02:03,505 INFO [train.py:901] (1/4) Epoch 24, batch 3550, loss[loss=0.208, simple_loss=0.3008, pruned_loss=0.05759, over 8321.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2878, pruned_loss=0.06146, over 1613173.51 frames. ], batch size: 25, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:02:26,769 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1407, 1.4478, 4.3669, 1.6698, 3.8385, 3.6434, 3.9544, 3.8333], + device='cuda:1'), covar=tensor([0.0664, 0.4744, 0.0492, 0.4385, 0.1101, 0.0943, 0.0615, 0.0714], + device='cuda:1'), in_proj_covar=tensor([0.0647, 0.0653, 0.0712, 0.0643, 0.0721, 0.0619, 0.0615, 0.0694], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:02:31,307 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.509e+02 2.981e+02 3.708e+02 7.370e+02, threshold=5.962e+02, percent-clipped=4.0 +2023-02-07 07:02:37,646 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189507.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:02:38,145 INFO [train.py:901] (1/4) Epoch 24, batch 3600, loss[loss=0.1835, simple_loss=0.2568, pruned_loss=0.05508, over 7572.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2874, pruned_loss=0.06132, over 1613958.86 frames. ], batch size: 18, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:02:45,843 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189519.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:03:01,838 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189541.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:03:12,170 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189556.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:03:13,411 INFO [train.py:901] (1/4) Epoch 24, batch 3650, loss[loss=0.1764, simple_loss=0.2566, pruned_loss=0.0481, over 7656.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2873, pruned_loss=0.0613, over 1613891.68 frames. ], batch size: 19, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:03:41,099 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 07:03:41,740 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 2.496e+02 2.930e+02 3.600e+02 6.319e+02, threshold=5.860e+02, percent-clipped=2.0 +2023-02-07 07:03:48,366 INFO [train.py:901] (1/4) Epoch 24, batch 3700, loss[loss=0.1992, simple_loss=0.2957, pruned_loss=0.0514, over 8353.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2869, pruned_loss=0.06058, over 1617286.90 frames. ], batch size: 24, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:03:49,803 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189610.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:23,135 INFO [train.py:901] (1/4) Epoch 24, batch 3750, loss[loss=0.1884, simple_loss=0.2692, pruned_loss=0.0538, over 7922.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2866, pruned_loss=0.06083, over 1613513.80 frames. ], batch size: 20, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:04:23,302 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189658.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:26,016 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189662.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:27,230 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189664.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:42,769 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189687.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:43,973 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189689.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:51,123 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.609e+02 3.129e+02 4.249e+02 7.016e+02, threshold=6.258e+02, percent-clipped=8.0 +2023-02-07 07:04:57,806 INFO [train.py:901] (1/4) Epoch 24, batch 3800, loss[loss=0.2017, simple_loss=0.2813, pruned_loss=0.06104, over 8488.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2863, pruned_loss=0.06061, over 1617132.28 frames. ], batch size: 28, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:04:58,620 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189709.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:07,484 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189722.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:09,548 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189725.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:24,362 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189746.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:32,061 INFO [train.py:901] (1/4) Epoch 24, batch 3850, loss[loss=0.2049, simple_loss=0.2982, pruned_loss=0.05577, over 8506.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2856, pruned_loss=0.06063, over 1607449.41 frames. ], batch size: 29, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:05:35,659 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189763.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:47,650 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 07:05:53,008 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189788.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:53,956 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 +2023-02-07 07:05:59,108 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189797.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:59,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.407e+02 2.910e+02 3.432e+02 8.251e+02, threshold=5.819e+02, percent-clipped=1.0 +2023-02-07 07:06:06,327 INFO [train.py:901] (1/4) Epoch 24, batch 3900, loss[loss=0.203, simple_loss=0.2819, pruned_loss=0.06206, over 8497.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2861, pruned_loss=0.06056, over 1609918.53 frames. ], batch size: 26, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:06:10,117 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189812.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:06:17,541 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189822.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:06:18,901 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189824.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:06:27,528 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189837.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:06:42,146 INFO [train.py:901] (1/4) Epoch 24, batch 3950, loss[loss=0.1711, simple_loss=0.2597, pruned_loss=0.04121, over 7972.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2853, pruned_loss=0.06007, over 1609243.28 frames. ], batch size: 21, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:06:45,540 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189863.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:07:09,592 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.387e+02 3.217e+02 3.997e+02 8.874e+02, threshold=6.434e+02, percent-clipped=5.0 +2023-02-07 07:07:16,323 INFO [train.py:901] (1/4) Epoch 24, batch 4000, loss[loss=0.1885, simple_loss=0.2846, pruned_loss=0.04616, over 8503.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2848, pruned_loss=0.05954, over 1611044.44 frames. ], batch size: 39, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:07:51,151 INFO [train.py:901] (1/4) Epoch 24, batch 4050, loss[loss=0.2444, simple_loss=0.3215, pruned_loss=0.08366, over 8351.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2851, pruned_loss=0.05988, over 1610120.59 frames. ], batch size: 26, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:08:05,461 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189978.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:08:07,509 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189981.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:08:18,681 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.334e+02 2.770e+02 3.399e+02 1.124e+03, threshold=5.539e+02, percent-clipped=1.0 +2023-02-07 07:08:22,559 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190002.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:08:26,235 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190006.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:08:27,386 INFO [train.py:901] (1/4) Epoch 24, batch 4100, loss[loss=0.1903, simple_loss=0.2877, pruned_loss=0.04645, over 8252.00 frames. ], tot_loss[loss=0.202, simple_loss=0.285, pruned_loss=0.05949, over 1612114.64 frames. ], batch size: 24, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:09:02,440 INFO [train.py:901] (1/4) Epoch 24, batch 4150, loss[loss=0.1915, simple_loss=0.2696, pruned_loss=0.05668, over 8098.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.285, pruned_loss=0.05926, over 1615671.71 frames. ], batch size: 21, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:09:08,094 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190066.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:09:17,894 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190080.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:09:25,159 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190090.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:09:30,579 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.686e+02 2.377e+02 2.724e+02 3.400e+02 7.023e+02, threshold=5.448e+02, percent-clipped=3.0 +2023-02-07 07:09:35,519 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190105.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:09:37,386 INFO [train.py:901] (1/4) Epoch 24, batch 4200, loss[loss=0.1719, simple_loss=0.2619, pruned_loss=0.04096, over 8291.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.285, pruned_loss=0.05957, over 1612461.12 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:09:43,686 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190117.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:09:45,031 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 +2023-02-07 07:09:48,195 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 07:09:50,485 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.98 vs. limit=5.0 +2023-02-07 07:10:10,661 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 07:10:11,425 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190157.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:10:11,871 INFO [train.py:901] (1/4) Epoch 24, batch 4250, loss[loss=0.2302, simple_loss=0.3254, pruned_loss=0.06746, over 8323.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2852, pruned_loss=0.05948, over 1615413.23 frames. ], batch size: 25, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:10:19,344 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190169.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:10:28,699 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190181.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:10:40,141 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.309e+02 2.865e+02 3.517e+02 8.092e+02, threshold=5.730e+02, percent-clipped=6.0 +2023-02-07 07:10:45,776 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190205.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:10:47,630 INFO [train.py:901] (1/4) Epoch 24, batch 4300, loss[loss=0.2178, simple_loss=0.3031, pruned_loss=0.06627, over 8438.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2852, pruned_loss=0.05983, over 1615453.92 frames. ], batch size: 27, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:10:56,467 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3404, 1.5725, 4.6240, 1.9684, 3.7588, 3.7205, 4.1939, 4.1173], + device='cuda:1'), covar=tensor([0.1224, 0.6515, 0.0991, 0.4882, 0.2074, 0.1472, 0.0993, 0.0986], + device='cuda:1'), in_proj_covar=tensor([0.0651, 0.0657, 0.0718, 0.0645, 0.0723, 0.0622, 0.0622, 0.0697], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:11:05,301 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190234.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:11:21,910 INFO [train.py:901] (1/4) Epoch 24, batch 4350, loss[loss=0.1735, simple_loss=0.2639, pruned_loss=0.04157, over 8247.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2855, pruned_loss=0.0599, over 1613896.80 frames. ], batch size: 22, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:11:22,815 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190259.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:11:40,177 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 07:11:50,408 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.430e+02 2.823e+02 3.493e+02 1.012e+03, threshold=5.646e+02, percent-clipped=3.0 +2023-02-07 07:11:57,334 INFO [train.py:901] (1/4) Epoch 24, batch 4400, loss[loss=0.2262, simple_loss=0.309, pruned_loss=0.07165, over 8191.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2867, pruned_loss=0.0605, over 1617381.57 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:12:15,563 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190334.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:12:23,134 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 07:12:32,982 INFO [train.py:901] (1/4) Epoch 24, batch 4450, loss[loss=0.2327, simple_loss=0.3, pruned_loss=0.08271, over 6922.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2854, pruned_loss=0.06016, over 1613741.49 frames. ], batch size: 71, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:12:43,427 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190373.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:00,241 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.201e+02 2.691e+02 3.403e+02 6.534e+02, threshold=5.381e+02, percent-clipped=2.0 +2023-02-07 07:13:00,469 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190398.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:07,768 INFO [train.py:901] (1/4) Epoch 24, batch 4500, loss[loss=0.2095, simple_loss=0.2962, pruned_loss=0.06144, over 8106.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2849, pruned_loss=0.0596, over 1616583.29 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:13:14,925 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190417.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:17,409 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 07:13:29,019 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190437.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:42,693 INFO [train.py:901] (1/4) Epoch 24, batch 4550, loss[loss=0.2056, simple_loss=0.2891, pruned_loss=0.06109, over 8499.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2847, pruned_loss=0.0598, over 1616416.25 frames. ], batch size: 28, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:13:44,873 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190461.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:45,495 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190462.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:47,435 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.9606, 2.3113, 3.6944, 1.9343, 1.7262, 3.6862, 0.6238, 2.1263], + device='cuda:1'), covar=tensor([0.1293, 0.1344, 0.0217, 0.1637, 0.2636, 0.0260, 0.2167, 0.1384], + device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0201, 0.0130, 0.0222, 0.0273, 0.0138, 0.0171, 0.0196], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 07:14:02,627 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190486.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:14:10,718 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.430e+02 2.973e+02 3.981e+02 9.647e+02, threshold=5.946e+02, percent-clipped=9.0 +2023-02-07 07:14:12,831 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190501.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:14:17,564 INFO [train.py:901] (1/4) Epoch 24, batch 4600, loss[loss=0.1766, simple_loss=0.2564, pruned_loss=0.04835, over 7641.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2841, pruned_loss=0.0596, over 1613866.18 frames. ], batch size: 19, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:14:21,248 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190513.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:14:42,211 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 +2023-02-07 07:14:54,248 INFO [train.py:901] (1/4) Epoch 24, batch 4650, loss[loss=0.1637, simple_loss=0.2449, pruned_loss=0.04119, over 7660.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2831, pruned_loss=0.05913, over 1609441.07 frames. ], batch size: 19, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:15:22,380 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.469e+02 2.987e+02 3.787e+02 1.231e+03, threshold=5.974e+02, percent-clipped=5.0 +2023-02-07 07:15:29,195 INFO [train.py:901] (1/4) Epoch 24, batch 4700, loss[loss=0.1909, simple_loss=0.2753, pruned_loss=0.05319, over 7971.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2821, pruned_loss=0.05872, over 1604837.71 frames. ], batch size: 21, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:15:29,946 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190609.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:15:34,427 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190616.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:15:41,739 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190627.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:15:42,471 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190628.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:16:04,016 INFO [train.py:901] (1/4) Epoch 24, batch 4750, loss[loss=0.2139, simple_loss=0.2949, pruned_loss=0.06648, over 8512.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2839, pruned_loss=0.05953, over 1608994.12 frames. ], batch size: 26, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:16:18,704 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190678.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:16:20,758 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 07:16:22,899 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 07:16:29,415 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190693.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:16:32,693 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.279e+02 2.789e+02 3.393e+02 7.815e+02, threshold=5.578e+02, percent-clipped=3.0 +2023-02-07 07:16:40,395 INFO [train.py:901] (1/4) Epoch 24, batch 4800, loss[loss=0.1747, simple_loss=0.2405, pruned_loss=0.05446, over 7428.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2831, pruned_loss=0.05858, over 1606704.00 frames. ], batch size: 17, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:17:07,852 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6995, 2.3063, 4.0264, 1.5132, 2.8760, 2.1361, 1.9408, 2.8332], + device='cuda:1'), covar=tensor([0.2190, 0.2677, 0.0969, 0.4988, 0.2088, 0.3593, 0.2526, 0.2651], + device='cuda:1'), in_proj_covar=tensor([0.0525, 0.0612, 0.0552, 0.0650, 0.0646, 0.0595, 0.0545, 0.0632], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:17:13,019 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 07:17:15,041 INFO [train.py:901] (1/4) Epoch 24, batch 4850, loss[loss=0.2218, simple_loss=0.3142, pruned_loss=0.06464, over 8101.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2828, pruned_loss=0.05845, over 1609699.97 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:17:17,937 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190761.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:17:27,638 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 +2023-02-07 07:17:40,143 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190793.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:17:43,507 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.278e+02 2.775e+02 3.178e+02 7.824e+02, threshold=5.550e+02, percent-clipped=3.0 +2023-02-07 07:17:50,724 INFO [train.py:901] (1/4) Epoch 24, batch 4900, loss[loss=0.218, simple_loss=0.3043, pruned_loss=0.0658, over 8738.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2834, pruned_loss=0.05867, over 1616782.27 frames. ], batch size: 39, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:18:06,940 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0463, 1.6747, 3.4548, 1.5271, 2.5076, 3.7527, 3.8686, 3.2303], + device='cuda:1'), covar=tensor([0.1161, 0.1770, 0.0328, 0.2191, 0.0963, 0.0238, 0.0520, 0.0538], + device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0324, 0.0285, 0.0316, 0.0315, 0.0272, 0.0429, 0.0302], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 07:18:16,606 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.33 vs. limit=5.0 +2023-02-07 07:18:25,853 INFO [train.py:901] (1/4) Epoch 24, batch 4950, loss[loss=0.1984, simple_loss=0.2724, pruned_loss=0.06216, over 7792.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.284, pruned_loss=0.0588, over 1613259.21 frames. ], batch size: 19, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:18:34,366 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 07:18:36,137 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190872.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:18:39,592 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190876.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:18:45,178 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190884.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:18:53,848 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190897.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:18:54,309 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.730e+02 2.393e+02 2.890e+02 3.701e+02 7.772e+02, threshold=5.780e+02, percent-clipped=4.0 +2023-02-07 07:19:01,817 INFO [train.py:901] (1/4) Epoch 24, batch 5000, loss[loss=0.1736, simple_loss=0.2675, pruned_loss=0.03992, over 8462.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2842, pruned_loss=0.059, over 1613322.24 frames. ], batch size: 27, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:19:02,714 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190909.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:19:33,480 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190953.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:19:36,594 INFO [train.py:901] (1/4) Epoch 24, batch 5050, loss[loss=0.1782, simple_loss=0.2673, pruned_loss=0.0446, over 8105.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2842, pruned_loss=0.05917, over 1614148.14 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:19:45,291 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190971.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:19:51,179 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 07:20:04,904 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.504e+02 2.521e+02 3.221e+02 4.359e+02 8.705e+02, threshold=6.442e+02, percent-clipped=11.0 +2023-02-07 07:20:11,616 INFO [train.py:901] (1/4) Epoch 24, batch 5100, loss[loss=0.2197, simple_loss=0.2982, pruned_loss=0.07058, over 8284.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2845, pruned_loss=0.05944, over 1613854.97 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:20:31,903 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=191037.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:20:37,437 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9907, 1.1741, 1.1517, 0.6956, 1.1600, 0.9581, 0.1206, 1.1297], + device='cuda:1'), covar=tensor([0.0456, 0.0373, 0.0363, 0.0548, 0.0479, 0.0951, 0.0821, 0.0338], + device='cuda:1'), in_proj_covar=tensor([0.0464, 0.0406, 0.0360, 0.0458, 0.0390, 0.0546, 0.0401, 0.0435], + device='cuda:1'), out_proj_covar=tensor([1.2377e-04, 1.0598e-04, 9.4450e-05, 1.2020e-04, 1.0260e-04, 1.5321e-04, + 1.0777e-04, 1.1456e-04], device='cuda:1') +2023-02-07 07:20:40,077 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191049.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:20:46,726 INFO [train.py:901] (1/4) Epoch 24, batch 5150, loss[loss=0.2231, simple_loss=0.3041, pruned_loss=0.07102, over 8656.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2843, pruned_loss=0.0595, over 1614918.35 frames. ], batch size: 34, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:20:53,681 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191068.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:20:57,803 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191074.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:05,749 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191086.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:05,787 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191086.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:13,676 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 2.286e+02 2.691e+02 3.617e+02 7.196e+02, threshold=5.383e+02, percent-clipped=2.0 +2023-02-07 07:21:20,844 INFO [train.py:901] (1/4) Epoch 24, batch 5200, loss[loss=0.1954, simple_loss=0.2837, pruned_loss=0.05356, over 8472.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2851, pruned_loss=0.05975, over 1618391.04 frames. ], batch size: 28, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:21:38,127 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191132.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:50,039 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 07:21:50,822 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191150.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:52,268 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191152.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:55,605 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191157.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:56,104 INFO [train.py:901] (1/4) Epoch 24, batch 5250, loss[loss=0.2098, simple_loss=0.2947, pruned_loss=0.06245, over 8900.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2865, pruned_loss=0.06042, over 1623712.82 frames. ], batch size: 40, lr: 3.14e-03, grad_scale: 8.0 +2023-02-07 07:22:04,650 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8796, 3.4933, 2.1957, 2.8553, 2.7730, 2.0167, 2.6905, 2.8820], + device='cuda:1'), covar=tensor([0.1739, 0.0405, 0.1300, 0.0791, 0.0722, 0.1505, 0.1135, 0.1128], + device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0234, 0.0337, 0.0309, 0.0299, 0.0341, 0.0346, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 07:22:25,866 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.418e+02 2.847e+02 3.981e+02 6.971e+02, threshold=5.694e+02, percent-clipped=11.0 +2023-02-07 07:22:28,316 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191203.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:22:31,600 INFO [train.py:901] (1/4) Epoch 24, batch 5300, loss[loss=0.1924, simple_loss=0.2823, pruned_loss=0.05121, over 8423.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2868, pruned_loss=0.0602, over 1623284.64 frames. ], batch size: 49, lr: 3.14e-03, grad_scale: 8.0 +2023-02-07 07:23:07,062 INFO [train.py:901] (1/4) Epoch 24, batch 5350, loss[loss=0.1989, simple_loss=0.2968, pruned_loss=0.05052, over 8292.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2856, pruned_loss=0.05961, over 1611714.30 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 8.0 +2023-02-07 07:23:36,917 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.294e+02 2.754e+02 3.342e+02 1.056e+03, threshold=5.508e+02, percent-clipped=2.0 +2023-02-07 07:23:42,379 INFO [train.py:901] (1/4) Epoch 24, batch 5400, loss[loss=0.2192, simple_loss=0.299, pruned_loss=0.06972, over 8246.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2864, pruned_loss=0.05972, over 1616886.43 frames. ], batch size: 24, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:23:53,142 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191324.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:24:05,750 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191342.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:24:11,127 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191349.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:24:17,035 INFO [train.py:901] (1/4) Epoch 24, batch 5450, loss[loss=0.1615, simple_loss=0.2416, pruned_loss=0.04073, over 7724.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2859, pruned_loss=0.05928, over 1618952.32 frames. ], batch size: 18, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:24:23,281 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191367.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:24:34,970 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191383.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:24:39,466 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 07:24:42,256 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4480, 1.7476, 1.7805, 1.1753, 1.8309, 1.3802, 0.4607, 1.6939], + device='cuda:1'), covar=tensor([0.0681, 0.0445, 0.0371, 0.0704, 0.0596, 0.1105, 0.1021, 0.0342], + device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0404, 0.0359, 0.0456, 0.0388, 0.0543, 0.0400, 0.0434], + device='cuda:1'), out_proj_covar=tensor([1.2354e-04, 1.0556e-04, 9.4182e-05, 1.1972e-04, 1.0197e-04, 1.5222e-04, + 1.0752e-04, 1.1447e-04], device='cuda:1') +2023-02-07 07:24:46,091 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 2.294e+02 2.951e+02 3.676e+02 7.135e+02, threshold=5.902e+02, percent-clipped=5.0 +2023-02-07 07:24:52,418 INFO [train.py:901] (1/4) Epoch 24, batch 5500, loss[loss=0.2058, simple_loss=0.2885, pruned_loss=0.06155, over 6990.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2851, pruned_loss=0.0592, over 1615364.33 frames. ], batch size: 72, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:24:52,652 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191408.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:25:07,820 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=191430.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:25:10,102 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191433.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:25:28,140 INFO [train.py:901] (1/4) Epoch 24, batch 5550, loss[loss=0.1732, simple_loss=0.2607, pruned_loss=0.0428, over 7804.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2836, pruned_loss=0.05802, over 1618948.73 frames. ], batch size: 20, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:25:29,825 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1260, 1.9591, 2.5115, 2.1116, 2.4393, 2.1869, 1.9868, 1.3871], + device='cuda:1'), covar=tensor([0.5379, 0.4664, 0.2052, 0.3922, 0.2676, 0.3289, 0.2027, 0.5375], + device='cuda:1'), in_proj_covar=tensor([0.0943, 0.0992, 0.0812, 0.0962, 0.1000, 0.0905, 0.0754, 0.0830], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 07:25:53,336 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=191494.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:25:57,930 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.420e+02 3.039e+02 3.989e+02 7.925e+02, threshold=6.078e+02, percent-clipped=5.0 +2023-02-07 07:26:03,388 INFO [train.py:901] (1/4) Epoch 24, batch 5600, loss[loss=0.2047, simple_loss=0.2917, pruned_loss=0.05884, over 8103.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2836, pruned_loss=0.05838, over 1615492.74 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:26:05,898 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-02-07 07:26:20,816 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1717, 1.6360, 4.2730, 1.8593, 2.3828, 4.8338, 4.9677, 4.1982], + device='cuda:1'), covar=tensor([0.1214, 0.1861, 0.0301, 0.2132, 0.1291, 0.0190, 0.0359, 0.0525], + device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0323, 0.0286, 0.0315, 0.0315, 0.0273, 0.0430, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 07:26:29,667 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191545.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:26:30,968 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=191547.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:26:38,439 INFO [train.py:901] (1/4) Epoch 24, batch 5650, loss[loss=0.1785, simple_loss=0.2682, pruned_loss=0.04441, over 7975.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2846, pruned_loss=0.05922, over 1616692.46 frames. ], batch size: 21, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:26:39,697 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-07 07:26:45,400 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 07:27:08,570 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.481e+02 2.901e+02 3.753e+02 9.237e+02, threshold=5.802e+02, percent-clipped=5.0 +2023-02-07 07:27:14,231 INFO [train.py:901] (1/4) Epoch 24, batch 5700, loss[loss=0.1729, simple_loss=0.2625, pruned_loss=0.04163, over 8089.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2836, pruned_loss=0.05904, over 1612526.28 frames. ], batch size: 21, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:27:14,757 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 07:27:15,041 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191609.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:27:21,337 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.87 vs. limit=5.0 +2023-02-07 07:27:22,489 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8173, 1.6753, 1.9329, 1.6508, 0.9777, 1.7186, 2.2208, 1.9876], + device='cuda:1'), covar=tensor([0.0446, 0.1203, 0.1623, 0.1381, 0.0610, 0.1426, 0.0636, 0.0617], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0189, 0.0160, 0.0100, 0.0163, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 07:27:49,745 INFO [train.py:901] (1/4) Epoch 24, batch 5750, loss[loss=0.1769, simple_loss=0.2749, pruned_loss=0.03949, over 8326.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2841, pruned_loss=0.05906, over 1614880.55 frames. ], batch size: 25, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:27:52,733 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191662.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:27:53,269 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 07:28:19,026 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.664e+02 2.394e+02 2.726e+02 3.524e+02 6.240e+02, threshold=5.452e+02, percent-clipped=4.0 +2023-02-07 07:28:25,092 INFO [train.py:901] (1/4) Epoch 24, batch 5800, loss[loss=0.2128, simple_loss=0.2743, pruned_loss=0.07568, over 7284.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2835, pruned_loss=0.05886, over 1614309.18 frames. ], batch size: 16, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:28:38,086 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=191727.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:28:39,805 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-07 07:28:40,460 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-02-07 07:28:59,640 INFO [train.py:901] (1/4) Epoch 24, batch 5850, loss[loss=0.1649, simple_loss=0.2494, pruned_loss=0.0402, over 7662.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2824, pruned_loss=0.05822, over 1616682.10 frames. ], batch size: 19, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:29:29,158 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.223e+02 2.821e+02 3.422e+02 9.012e+02, threshold=5.641e+02, percent-clipped=8.0 +2023-02-07 07:29:30,138 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191801.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:29:34,718 INFO [train.py:901] (1/4) Epoch 24, batch 5900, loss[loss=0.1801, simple_loss=0.2707, pruned_loss=0.04477, over 8022.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2825, pruned_loss=0.0582, over 1613787.06 frames. ], batch size: 22, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:29:48,211 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191826.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:29:59,313 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191842.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:30:10,941 INFO [train.py:901] (1/4) Epoch 24, batch 5950, loss[loss=0.2098, simple_loss=0.2941, pruned_loss=0.06271, over 8460.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2837, pruned_loss=0.05849, over 1617641.38 frames. ], batch size: 27, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:30:16,037 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:30:33,612 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191890.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:30:40,234 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 2.338e+02 2.991e+02 3.628e+02 7.270e+02, threshold=5.982e+02, percent-clipped=3.0 +2023-02-07 07:30:45,682 INFO [train.py:901] (1/4) Epoch 24, batch 6000, loss[loss=0.1955, simple_loss=0.2703, pruned_loss=0.06036, over 7292.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.282, pruned_loss=0.0581, over 1612255.21 frames. ], batch size: 16, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:30:45,682 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 07:31:00,540 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7029, 1.4821, 3.8395, 1.4315, 3.4329, 3.1293, 3.5175, 3.3386], + device='cuda:1'), covar=tensor([0.0667, 0.4878, 0.0591, 0.4679, 0.1243, 0.1135, 0.0691, 0.0831], + device='cuda:1'), in_proj_covar=tensor([0.0648, 0.0653, 0.0710, 0.0643, 0.0720, 0.0617, 0.0618, 0.0690], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:31:01,027 INFO [train.py:935] (1/4) Epoch 24, validation: loss=0.1718, simple_loss=0.2718, pruned_loss=0.0359, over 944034.00 frames. +2023-02-07 07:31:01,027 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 07:31:08,203 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191918.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:31:12,826 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2074, 1.5454, 1.7644, 1.4692, 1.0409, 1.6401, 1.8972, 1.8202], + device='cuda:1'), covar=tensor([0.0508, 0.1253, 0.1704, 0.1450, 0.0606, 0.1424, 0.0670, 0.0581], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0189, 0.0160, 0.0101, 0.0163, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 07:31:24,853 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191943.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:31:26,286 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6506, 2.0283, 3.1753, 1.4760, 2.4785, 2.0795, 1.7156, 2.4500], + device='cuda:1'), covar=tensor([0.2150, 0.2923, 0.1001, 0.5024, 0.2158, 0.3516, 0.2662, 0.2596], + device='cuda:1'), in_proj_covar=tensor([0.0531, 0.0615, 0.0555, 0.0652, 0.0652, 0.0600, 0.0547, 0.0637], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:31:35,325 INFO [train.py:901] (1/4) Epoch 24, batch 6050, loss[loss=0.2113, simple_loss=0.3063, pruned_loss=0.0581, over 8295.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2831, pruned_loss=0.05827, over 1614115.83 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:31:41,188 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4854, 1.8354, 2.6054, 1.3831, 1.9220, 1.8565, 1.5954, 1.9567], + device='cuda:1'), covar=tensor([0.2015, 0.2604, 0.0965, 0.4843, 0.2068, 0.3382, 0.2560, 0.2422], + device='cuda:1'), in_proj_covar=tensor([0.0531, 0.0615, 0.0555, 0.0652, 0.0652, 0.0600, 0.0548, 0.0636], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:31:57,329 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7650, 1.3417, 1.6391, 1.2582, 0.9229, 1.4125, 1.6268, 1.4682], + device='cuda:1'), covar=tensor([0.0576, 0.1347, 0.1730, 0.1513, 0.0640, 0.1558, 0.0737, 0.0691], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0189, 0.0159, 0.0100, 0.0162, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 07:31:58,022 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5707, 3.0058, 2.5429, 4.2003, 1.8694, 2.1293, 2.8164, 3.0089], + device='cuda:1'), covar=tensor([0.0771, 0.0873, 0.0792, 0.0228, 0.1074, 0.1284, 0.0870, 0.0830], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0199, 0.0245, 0.0216, 0.0205, 0.0247, 0.0253, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 07:32:04,602 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.415e+02 2.742e+02 3.441e+02 8.508e+02, threshold=5.485e+02, percent-clipped=2.0 +2023-02-07 07:32:10,738 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7944, 2.4022, 4.1471, 1.5920, 3.2112, 2.3060, 1.9628, 3.0112], + device='cuda:1'), covar=tensor([0.2043, 0.2594, 0.0825, 0.4710, 0.1720, 0.3427, 0.2457, 0.2348], + device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0613, 0.0552, 0.0649, 0.0649, 0.0598, 0.0545, 0.0634], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:32:11,923 INFO [train.py:901] (1/4) Epoch 24, batch 6100, loss[loss=0.1813, simple_loss=0.2778, pruned_loss=0.04245, over 8332.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2827, pruned_loss=0.05773, over 1615083.45 frames. ], batch size: 26, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:32:32,943 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0826, 3.1975, 2.2114, 2.6931, 2.5332, 1.9393, 2.4990, 2.8632], + device='cuda:1'), covar=tensor([0.1461, 0.0389, 0.1126, 0.0698, 0.0661, 0.1436, 0.1064, 0.0941], + device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0234, 0.0338, 0.0311, 0.0302, 0.0343, 0.0348, 0.0320], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 07:32:34,059 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 07:32:37,032 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 +2023-02-07 07:32:46,755 INFO [train.py:901] (1/4) Epoch 24, batch 6150, loss[loss=0.2172, simple_loss=0.3053, pruned_loss=0.06452, over 8188.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.282, pruned_loss=0.05756, over 1613000.88 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:33:06,709 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 07:33:15,571 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192098.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:33:16,711 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.367e+02 2.762e+02 3.348e+02 6.106e+02, threshold=5.524e+02, percent-clipped=2.0 +2023-02-07 07:33:22,018 INFO [train.py:901] (1/4) Epoch 24, batch 6200, loss[loss=0.2249, simple_loss=0.3084, pruned_loss=0.07068, over 8350.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2828, pruned_loss=0.05863, over 1612089.15 frames. ], batch size: 26, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:33:30,650 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192120.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:33:32,770 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192123.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:33:50,993 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 07:33:56,607 INFO [train.py:901] (1/4) Epoch 24, batch 6250, loss[loss=0.1683, simple_loss=0.2465, pruned_loss=0.04502, over 7919.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2831, pruned_loss=0.05889, over 1612407.18 frames. ], batch size: 20, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:34:07,847 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192173.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:34:26,652 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.364e+02 2.949e+02 3.646e+02 8.976e+02, threshold=5.898e+02, percent-clipped=7.0 +2023-02-07 07:34:31,430 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 +2023-02-07 07:34:33,021 INFO [train.py:901] (1/4) Epoch 24, batch 6300, loss[loss=0.2407, simple_loss=0.3314, pruned_loss=0.07494, over 8513.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2828, pruned_loss=0.05907, over 1606651.67 frames. ], batch size: 26, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:34:52,688 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192237.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:34:54,069 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:35:02,879 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8483, 1.4822, 3.5234, 1.5702, 2.5196, 3.9020, 4.0212, 3.3977], + device='cuda:1'), covar=tensor([0.1250, 0.1847, 0.0330, 0.2045, 0.1048, 0.0207, 0.0504, 0.0542], + device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0321, 0.0284, 0.0312, 0.0312, 0.0270, 0.0427, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 07:35:07,367 INFO [train.py:901] (1/4) Epoch 24, batch 6350, loss[loss=0.2422, simple_loss=0.3165, pruned_loss=0.08392, over 6577.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2824, pruned_loss=0.05861, over 1608090.13 frames. ], batch size: 71, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:35:26,356 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-07 07:35:36,835 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.340e+02 2.329e+02 2.896e+02 3.640e+02 5.459e+02, threshold=5.791e+02, percent-clipped=0.0 +2023-02-07 07:35:43,003 INFO [train.py:901] (1/4) Epoch 24, batch 6400, loss[loss=0.214, simple_loss=0.3079, pruned_loss=0.06011, over 8019.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2822, pruned_loss=0.0586, over 1607740.61 frames. ], batch size: 22, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:36:19,126 INFO [train.py:901] (1/4) Epoch 24, batch 6450, loss[loss=0.1954, simple_loss=0.2758, pruned_loss=0.05752, over 7968.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2835, pruned_loss=0.05942, over 1610887.76 frames. ], batch size: 21, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:36:22,558 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-07 07:36:49,106 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.491e+02 2.965e+02 3.858e+02 7.678e+02, threshold=5.930e+02, percent-clipped=7.0 +2023-02-07 07:36:54,683 INFO [train.py:901] (1/4) Epoch 24, batch 6500, loss[loss=0.2113, simple_loss=0.3087, pruned_loss=0.05692, over 8474.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2849, pruned_loss=0.0602, over 1610641.99 frames. ], batch size: 29, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:36:56,151 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192410.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:37:00,679 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192417.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:37:25,023 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192451.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:37:29,459 INFO [train.py:901] (1/4) Epoch 24, batch 6550, loss[loss=0.2246, simple_loss=0.3072, pruned_loss=0.07094, over 8340.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2854, pruned_loss=0.05995, over 1614043.37 frames. ], batch size: 25, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:37:33,653 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192464.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:37:48,312 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 07:37:53,276 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7409, 2.1181, 3.2200, 1.5628, 2.5716, 2.0853, 1.8685, 2.5923], + device='cuda:1'), covar=tensor([0.1874, 0.2580, 0.0769, 0.4479, 0.1682, 0.3150, 0.2271, 0.2125], + device='cuda:1'), in_proj_covar=tensor([0.0533, 0.0619, 0.0558, 0.0656, 0.0654, 0.0603, 0.0550, 0.0639], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:37:58,368 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.233e+02 2.734e+02 3.455e+02 6.558e+02, threshold=5.467e+02, percent-clipped=2.0 +2023-02-07 07:38:03,875 INFO [train.py:901] (1/4) Epoch 24, batch 6600, loss[loss=0.1938, simple_loss=0.2712, pruned_loss=0.05825, over 7654.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2843, pruned_loss=0.05933, over 1615508.65 frames. ], batch size: 19, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:38:08,120 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 07:38:10,835 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:38:24,030 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 +2023-02-07 07:38:39,631 INFO [train.py:901] (1/4) Epoch 24, batch 6650, loss[loss=0.1764, simple_loss=0.2713, pruned_loss=0.04078, over 8188.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2842, pruned_loss=0.05919, over 1614135.52 frames. ], batch size: 23, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:38:53,942 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192579.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:38:55,127 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192581.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:38:56,510 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192583.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:38:56,613 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6907, 1.9798, 2.0391, 1.8889, 1.3187, 1.9715, 2.4134, 2.0991], + device='cuda:1'), covar=tensor([0.0497, 0.1070, 0.1506, 0.1222, 0.0599, 0.1225, 0.0607, 0.0525], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0189, 0.0159, 0.0100, 0.0162, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 07:39:08,643 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.608e+02 2.365e+02 2.876e+02 3.746e+02 9.522e+02, threshold=5.752e+02, percent-clipped=3.0 +2023-02-07 07:39:14,185 INFO [train.py:901] (1/4) Epoch 24, batch 6700, loss[loss=0.1704, simple_loss=0.2677, pruned_loss=0.03654, over 8354.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2843, pruned_loss=0.05934, over 1611603.41 frames. ], batch size: 24, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:39:31,226 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192632.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:39:32,527 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192634.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:39:49,023 INFO [train.py:901] (1/4) Epoch 24, batch 6750, loss[loss=0.1662, simple_loss=0.251, pruned_loss=0.04067, over 7808.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2842, pruned_loss=0.05937, over 1613057.28 frames. ], batch size: 20, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:39:59,811 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3085, 1.5237, 4.4256, 1.8898, 2.5909, 5.1359, 5.2080, 4.5116], + device='cuda:1'), covar=tensor([0.1124, 0.1922, 0.0272, 0.1922, 0.1152, 0.0163, 0.0393, 0.0504], + device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0323, 0.0287, 0.0315, 0.0314, 0.0273, 0.0430, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 07:40:01,189 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3883, 2.3606, 1.9552, 2.2741, 2.0798, 1.5773, 1.9435, 1.9104], + device='cuda:1'), covar=tensor([0.1421, 0.0426, 0.1185, 0.0533, 0.0744, 0.1619, 0.0992, 0.0942], + device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0235, 0.0336, 0.0307, 0.0301, 0.0341, 0.0346, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 07:40:12,032 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192691.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:40:15,248 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192696.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:40:16,605 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192698.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:40:17,824 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.601e+02 3.163e+02 3.841e+02 9.507e+02, threshold=6.325e+02, percent-clipped=3.0 +2023-02-07 07:40:21,125 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 07:40:23,402 INFO [train.py:901] (1/4) Epoch 24, batch 6800, loss[loss=0.2355, simple_loss=0.3184, pruned_loss=0.0763, over 8576.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2838, pruned_loss=0.05943, over 1607458.44 frames. ], batch size: 31, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:40:24,828 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 07:40:51,292 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2654, 1.9637, 2.5271, 1.3584, 1.5959, 2.5452, 1.3306, 2.0233], + device='cuda:1'), covar=tensor([0.1398, 0.0935, 0.0412, 0.1436, 0.1853, 0.0396, 0.1438, 0.1274], + device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0199, 0.0129, 0.0220, 0.0270, 0.0138, 0.0171, 0.0196], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 07:40:56,190 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192753.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:40:56,803 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192754.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:40:59,526 INFO [train.py:901] (1/4) Epoch 24, batch 6850, loss[loss=0.1999, simple_loss=0.2844, pruned_loss=0.05772, over 8200.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2837, pruned_loss=0.05927, over 1610259.77 frames. ], batch size: 23, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:41:01,649 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192761.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:41:15,304 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 07:41:26,178 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192795.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:41:29,430 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.230e+02 2.864e+02 3.611e+02 9.090e+02, threshold=5.729e+02, percent-clipped=1.0 +2023-02-07 07:41:35,201 INFO [train.py:901] (1/4) Epoch 24, batch 6900, loss[loss=0.2263, simple_loss=0.2948, pruned_loss=0.07891, over 6814.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2845, pruned_loss=0.05948, over 1611590.41 frames. ], batch size: 15, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:41:54,067 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192835.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:03,108 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192849.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:42:08,879 INFO [train.py:901] (1/4) Epoch 24, batch 6950, loss[loss=0.2089, simple_loss=0.2798, pruned_loss=0.06896, over 7699.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2853, pruned_loss=0.06025, over 1609436.54 frames. ], batch size: 18, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:42:10,326 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192860.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:17,133 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192869.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:21,981 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192876.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:23,184 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 07:42:23,754 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-07 07:42:30,036 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192888.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:38,586 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.384e+02 2.955e+02 3.597e+02 9.319e+02, threshold=5.910e+02, percent-clipped=1.0 +2023-02-07 07:42:44,709 INFO [train.py:901] (1/4) Epoch 24, batch 7000, loss[loss=0.2065, simple_loss=0.2945, pruned_loss=0.05928, over 8251.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2846, pruned_loss=0.05943, over 1613657.32 frames. ], batch size: 24, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:42:46,242 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192910.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:48,264 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192913.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:48,881 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([6.0364, 1.8109, 6.2124, 2.2274, 5.6335, 5.2709, 5.7439, 5.6380], + device='cuda:1'), covar=tensor([0.0401, 0.4282, 0.0284, 0.3828, 0.0827, 0.0720, 0.0403, 0.0458], + device='cuda:1'), in_proj_covar=tensor([0.0641, 0.0651, 0.0707, 0.0637, 0.0712, 0.0612, 0.0614, 0.0683], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:43:15,544 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192952.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:16,861 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192954.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:19,400 INFO [train.py:901] (1/4) Epoch 24, batch 7050, loss[loss=0.1557, simple_loss=0.2438, pruned_loss=0.03377, over 7684.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2844, pruned_loss=0.05929, over 1612676.26 frames. ], batch size: 18, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:43:32,663 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192977.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:33,235 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192978.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:34,071 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192979.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:46,466 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4676, 1.3894, 1.7421, 1.1840, 1.1355, 1.7278, 0.2723, 1.1766], + device='cuda:1'), covar=tensor([0.1680, 0.1231, 0.0394, 0.0903, 0.2579, 0.0438, 0.1965, 0.1334], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0200, 0.0130, 0.0221, 0.0273, 0.0139, 0.0171, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 07:43:48,981 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 2.407e+02 3.090e+02 3.925e+02 9.689e+02, threshold=6.179e+02, percent-clipped=7.0 +2023-02-07 07:43:54,448 INFO [train.py:901] (1/4) Epoch 24, batch 7100, loss[loss=0.2186, simple_loss=0.3082, pruned_loss=0.06449, over 8467.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.286, pruned_loss=0.05959, over 1614289.11 frames. ], batch size: 25, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:44:14,218 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193035.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:44:15,044 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8578, 1.7947, 2.4840, 1.5904, 1.4236, 2.4513, 0.5670, 1.5197], + device='cuda:1'), covar=tensor([0.1735, 0.1258, 0.0314, 0.1207, 0.2584, 0.0359, 0.2046, 0.1333], + device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0200, 0.0129, 0.0221, 0.0273, 0.0139, 0.0171, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 07:44:29,518 INFO [train.py:901] (1/4) Epoch 24, batch 7150, loss[loss=0.1881, simple_loss=0.2666, pruned_loss=0.0548, over 8083.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2851, pruned_loss=0.05911, over 1612915.37 frames. ], batch size: 21, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:44:30,370 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193059.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:44:54,300 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193093.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:44:56,928 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193097.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:44:58,920 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.272e+02 2.945e+02 3.915e+02 7.728e+02, threshold=5.890e+02, percent-clipped=4.0 +2023-02-07 07:45:05,024 INFO [train.py:901] (1/4) Epoch 24, batch 7200, loss[loss=0.1886, simple_loss=0.2788, pruned_loss=0.04924, over 8470.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2853, pruned_loss=0.05901, over 1614127.51 frames. ], batch size: 25, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:45:16,871 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193125.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:45:21,519 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193132.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:45:34,270 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193150.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:45:34,293 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193150.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:45:39,614 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193157.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:45:40,079 INFO [train.py:901] (1/4) Epoch 24, batch 7250, loss[loss=0.2026, simple_loss=0.2881, pruned_loss=0.05853, over 8469.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.285, pruned_loss=0.05907, over 1612705.59 frames. ], batch size: 27, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:45:45,716 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193166.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:46:03,181 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193191.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:46:04,337 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193193.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:46:08,901 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 2.389e+02 2.780e+02 3.377e+02 1.311e+03, threshold=5.561e+02, percent-clipped=2.0 +2023-02-07 07:46:14,404 INFO [train.py:901] (1/4) Epoch 24, batch 7300, loss[loss=0.2146, simple_loss=0.2902, pruned_loss=0.0695, over 8133.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2863, pruned_loss=0.05975, over 1614715.71 frames. ], batch size: 22, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:46:17,193 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193212.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:46:29,142 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193229.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:46:43,419 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-07 07:46:48,080 INFO [train.py:901] (1/4) Epoch 24, batch 7350, loss[loss=0.1702, simple_loss=0.2497, pruned_loss=0.04532, over 7233.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2859, pruned_loss=0.05978, over 1614592.75 frames. ], batch size: 16, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:47:00,601 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 07:47:11,639 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 07:47:17,800 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.477e+02 2.971e+02 3.853e+02 6.522e+02, threshold=5.942e+02, percent-clipped=4.0 +2023-02-07 07:47:19,302 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193302.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:47:23,214 INFO [train.py:901] (1/4) Epoch 24, batch 7400, loss[loss=0.193, simple_loss=0.2609, pruned_loss=0.06255, over 7234.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2846, pruned_loss=0.05926, over 1611308.93 frames. ], batch size: 16, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:47:23,404 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193308.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:47:31,926 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 07:47:52,386 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193349.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:47:58,367 INFO [train.py:901] (1/4) Epoch 24, batch 7450, loss[loss=0.1864, simple_loss=0.2772, pruned_loss=0.04773, over 8641.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2846, pruned_loss=0.05954, over 1611257.47 frames. ], batch size: 49, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:48:09,543 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 07:48:10,327 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193374.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:48:21,035 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2079, 4.1901, 3.7539, 1.8193, 3.7290, 3.8071, 3.6524, 3.6148], + device='cuda:1'), covar=tensor([0.0784, 0.0584, 0.1080, 0.5124, 0.0935, 0.1039, 0.1401, 0.0815], + device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0455, 0.0443, 0.0557, 0.0442, 0.0456, 0.0437, 0.0397], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:48:26,201 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.13 vs. limit=5.0 +2023-02-07 07:48:28,445 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.340e+02 2.930e+02 4.048e+02 8.147e+02, threshold=5.861e+02, percent-clipped=5.0 +2023-02-07 07:48:30,641 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193403.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:48:32,909 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193406.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:48:34,038 INFO [train.py:901] (1/4) Epoch 24, batch 7500, loss[loss=0.1812, simple_loss=0.2668, pruned_loss=0.04785, over 7662.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2875, pruned_loss=0.06143, over 1615713.30 frames. ], batch size: 19, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:48:50,723 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193431.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:48:56,215 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193439.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:49:09,712 INFO [train.py:901] (1/4) Epoch 24, batch 7550, loss[loss=0.1911, simple_loss=0.2678, pruned_loss=0.05718, over 7445.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.285, pruned_loss=0.06034, over 1607510.28 frames. ], batch size: 17, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:49:16,841 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193468.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:49:18,144 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9168, 1.6206, 3.4552, 1.5632, 2.3542, 3.7826, 3.9825, 3.2065], + device='cuda:1'), covar=tensor([0.1302, 0.1781, 0.0326, 0.2094, 0.1158, 0.0247, 0.0510, 0.0525], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0322, 0.0286, 0.0314, 0.0314, 0.0272, 0.0428, 0.0302], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 07:49:19,406 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193472.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:49:32,473 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.89 vs. limit=5.0 +2023-02-07 07:49:34,405 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193493.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:49:39,067 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.474e+02 3.046e+02 3.751e+02 6.843e+02, threshold=6.092e+02, percent-clipped=3.0 +2023-02-07 07:49:45,271 INFO [train.py:901] (1/4) Epoch 24, batch 7600, loss[loss=0.208, simple_loss=0.288, pruned_loss=0.06402, over 7129.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2851, pruned_loss=0.06076, over 1608649.71 frames. ], batch size: 72, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:49:51,993 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193518.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:50:19,354 INFO [train.py:901] (1/4) Epoch 24, batch 7650, loss[loss=0.1574, simple_loss=0.2398, pruned_loss=0.03752, over 7818.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2855, pruned_loss=0.06132, over 1609216.98 frames. ], batch size: 20, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:50:24,424 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193564.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:50:30,315 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193573.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:50:41,144 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193589.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:50:48,583 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.797e+02 2.626e+02 3.196e+02 4.372e+02 7.437e+02, threshold=6.392e+02, percent-clipped=4.0 +2023-02-07 07:50:53,961 INFO [train.py:901] (1/4) Epoch 24, batch 7700, loss[loss=0.2617, simple_loss=0.3296, pruned_loss=0.09685, over 7340.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2867, pruned_loss=0.06139, over 1610348.22 frames. ], batch size: 71, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:51:11,796 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193633.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:51:14,902 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 07:51:20,931 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193646.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:51:27,965 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1090, 1.7908, 2.3029, 1.9376, 2.2379, 2.1382, 1.9937, 1.1300], + device='cuda:1'), covar=tensor([0.5424, 0.4907, 0.1991, 0.3916, 0.2524, 0.3081, 0.1966, 0.5146], + device='cuda:1'), in_proj_covar=tensor([0.0950, 0.1001, 0.0824, 0.0969, 0.1011, 0.0912, 0.0763, 0.0836], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 07:51:29,074 INFO [train.py:901] (1/4) Epoch 24, batch 7750, loss[loss=0.2162, simple_loss=0.3016, pruned_loss=0.06544, over 8501.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2873, pruned_loss=0.06139, over 1614562.83 frames. ], batch size: 26, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:51:50,184 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193688.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:51:58,105 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.447e+02 2.679e+02 3.147e+02 3.999e+02 8.742e+02, threshold=6.294e+02, percent-clipped=3.0 +2023-02-07 07:52:03,353 INFO [train.py:901] (1/4) Epoch 24, batch 7800, loss[loss=0.2185, simple_loss=0.3046, pruned_loss=0.06621, over 8475.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2867, pruned_loss=0.06096, over 1613480.31 frames. ], batch size: 29, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:52:37,263 INFO [train.py:901] (1/4) Epoch 24, batch 7850, loss[loss=0.2018, simple_loss=0.2743, pruned_loss=0.06467, over 7643.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2866, pruned_loss=0.06069, over 1616428.93 frames. ], batch size: 19, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:52:39,491 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193761.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:52:47,566 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5014, 2.0633, 3.2613, 1.4357, 2.4362, 1.9751, 1.6111, 2.4940], + device='cuda:1'), covar=tensor([0.2159, 0.2810, 0.0910, 0.4846, 0.2118, 0.3475, 0.2652, 0.2367], + device='cuda:1'), in_proj_covar=tensor([0.0530, 0.0617, 0.0556, 0.0651, 0.0651, 0.0599, 0.0548, 0.0636], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 07:52:48,222 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193774.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:52:54,181 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193783.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:53:05,055 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193799.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:53:05,511 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.471e+02 2.801e+02 3.652e+02 8.352e+02, threshold=5.603e+02, percent-clipped=2.0 +2023-02-07 07:53:10,835 INFO [train.py:901] (1/4) Epoch 24, batch 7900, loss[loss=0.1746, simple_loss=0.2647, pruned_loss=0.04227, over 7808.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2866, pruned_loss=0.06082, over 1619135.31 frames. ], batch size: 20, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:53:16,267 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193816.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:53:23,768 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.28 vs. limit=5.0 +2023-02-07 07:53:28,183 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2228, 2.1046, 1.6358, 1.9561, 1.7377, 1.3948, 1.5921, 1.6346], + device='cuda:1'), covar=tensor([0.1421, 0.0444, 0.1366, 0.0590, 0.0795, 0.1751, 0.1121, 0.0976], + device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0236, 0.0341, 0.0311, 0.0303, 0.0345, 0.0350, 0.0322], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 07:53:43,830 INFO [train.py:901] (1/4) Epoch 24, batch 7950, loss[loss=0.1978, simple_loss=0.2795, pruned_loss=0.05805, over 8346.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2875, pruned_loss=0.06143, over 1617817.79 frames. ], batch size: 26, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:54:11,269 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193898.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:54:12,511 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.494e+02 3.061e+02 3.521e+02 6.741e+02, threshold=6.122e+02, percent-clipped=2.0 +2023-02-07 07:54:13,969 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193902.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:54:17,826 INFO [train.py:901] (1/4) Epoch 24, batch 8000, loss[loss=0.1941, simple_loss=0.2708, pruned_loss=0.05869, over 7652.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2877, pruned_loss=0.06127, over 1619287.30 frames. ], batch size: 19, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:54:33,384 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193931.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:54:42,139 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193944.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:54:43,444 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0352, 2.1981, 1.8464, 2.4918, 1.6865, 1.7896, 1.9775, 2.2269], + device='cuda:1'), covar=tensor([0.0653, 0.0632, 0.0819, 0.0450, 0.0893, 0.1061, 0.0709, 0.0672], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0197, 0.0245, 0.0214, 0.0205, 0.0247, 0.0251, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 07:54:51,055 INFO [train.py:901] (1/4) Epoch 24, batch 8050, loss[loss=0.2318, simple_loss=0.3067, pruned_loss=0.07849, over 7304.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2861, pruned_loss=0.06127, over 1600025.46 frames. ], batch size: 75, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:54:58,127 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193969.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:55:03,440 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193977.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:55:23,284 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 07:55:28,455 INFO [train.py:901] (1/4) Epoch 25, batch 0, loss[loss=0.2415, simple_loss=0.3022, pruned_loss=0.09038, over 8247.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3022, pruned_loss=0.09038, over 8247.00 frames. ], batch size: 24, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:55:28,456 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 07:55:39,676 INFO [train.py:935] (1/4) Epoch 25, validation: loss=0.1722, simple_loss=0.2724, pruned_loss=0.03604, over 944034.00 frames. +2023-02-07 07:55:39,677 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 07:55:46,485 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.577e+02 3.086e+02 3.975e+02 9.885e+02, threshold=6.172e+02, percent-clipped=3.0 +2023-02-07 07:55:57,042 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 07:56:00,101 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194017.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:56:15,956 INFO [train.py:901] (1/4) Epoch 25, batch 50, loss[loss=0.2663, simple_loss=0.3293, pruned_loss=0.1016, over 6930.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2846, pruned_loss=0.05953, over 364272.69 frames. ], batch size: 72, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:56:17,544 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194042.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:56:32,521 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 07:56:51,135 INFO [train.py:901] (1/4) Epoch 25, batch 100, loss[loss=0.1649, simple_loss=0.2509, pruned_loss=0.0394, over 7420.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2852, pruned_loss=0.05933, over 643337.17 frames. ], batch size: 17, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:56:51,286 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194090.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:56:52,672 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194092.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:56:55,700 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 07:56:57,728 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.655e+02 3.251e+02 4.247e+02 7.218e+02, threshold=6.502e+02, percent-clipped=2.0 +2023-02-07 07:57:19,065 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 07:57:22,495 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 07:57:25,373 INFO [train.py:901] (1/4) Epoch 25, batch 150, loss[loss=0.1951, simple_loss=0.2754, pruned_loss=0.05736, over 7968.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.286, pruned_loss=0.06034, over 863209.20 frames. ], batch size: 21, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:57:35,092 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194154.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:57:52,133 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194179.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:57:58,197 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194187.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:58:00,647 INFO [train.py:901] (1/4) Epoch 25, batch 200, loss[loss=0.2202, simple_loss=0.3068, pruned_loss=0.06681, over 8444.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2858, pruned_loss=0.05953, over 1033325.69 frames. ], batch size: 29, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:58:07,397 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.663e+02 2.343e+02 2.842e+02 3.543e+02 5.999e+02, threshold=5.685e+02, percent-clipped=0.0 +2023-02-07 07:58:16,692 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194212.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:58:35,370 INFO [train.py:901] (1/4) Epoch 25, batch 250, loss[loss=0.2108, simple_loss=0.2988, pruned_loss=0.06146, over 8373.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2865, pruned_loss=0.05968, over 1167797.73 frames. ], batch size: 24, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:58:39,450 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194246.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:58:49,469 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 07:58:58,139 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 07:59:08,685 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-07 07:59:09,584 INFO [train.py:901] (1/4) Epoch 25, batch 300, loss[loss=0.1614, simple_loss=0.2475, pruned_loss=0.03766, over 7695.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2848, pruned_loss=0.05898, over 1265161.46 frames. ], batch size: 18, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:59:17,103 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.353e+02 2.857e+02 3.504e+02 7.851e+02, threshold=5.715e+02, percent-clipped=2.0 +2023-02-07 07:59:19,404 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7572, 1.5307, 3.2280, 1.3998, 2.3856, 3.3932, 3.5757, 2.9145], + device='cuda:1'), covar=tensor([0.1278, 0.1664, 0.0304, 0.2179, 0.0856, 0.0267, 0.0571, 0.0535], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0322, 0.0287, 0.0316, 0.0316, 0.0273, 0.0429, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 07:59:43,419 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194336.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:59:45,908 INFO [train.py:901] (1/4) Epoch 25, batch 350, loss[loss=0.1967, simple_loss=0.2797, pruned_loss=0.05687, over 7544.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2851, pruned_loss=0.05938, over 1341544.82 frames. ], batch size: 18, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:59:51,517 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194348.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:00:00,445 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194361.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:00:08,055 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194371.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:00:09,434 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194373.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:00:14,131 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194380.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:00:20,654 INFO [train.py:901] (1/4) Epoch 25, batch 400, loss[loss=0.2143, simple_loss=0.2858, pruned_loss=0.07137, over 7232.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2859, pruned_loss=0.05995, over 1401464.36 frames. ], batch size: 16, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 08:00:27,619 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.445e+02 3.013e+02 3.982e+02 8.525e+02, threshold=6.027e+02, percent-clipped=7.0 +2023-02-07 08:00:29,489 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 +2023-02-07 08:00:52,196 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194434.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:00:56,970 INFO [train.py:901] (1/4) Epoch 25, batch 450, loss[loss=0.2734, simple_loss=0.3418, pruned_loss=0.1025, over 7217.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2845, pruned_loss=0.05904, over 1448401.78 frames. ], batch size: 73, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 08:01:30,918 INFO [train.py:901] (1/4) Epoch 25, batch 500, loss[loss=0.1937, simple_loss=0.2678, pruned_loss=0.0598, over 7283.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2859, pruned_loss=0.05997, over 1487048.17 frames. ], batch size: 16, lr: 3.04e-03, grad_scale: 16.0 +2023-02-07 08:01:37,839 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.459e+02 3.156e+02 4.025e+02 7.800e+02, threshold=6.312e+02, percent-clipped=3.0 +2023-02-07 08:02:06,155 INFO [train.py:901] (1/4) Epoch 25, batch 550, loss[loss=0.1734, simple_loss=0.2557, pruned_loss=0.04552, over 7660.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2851, pruned_loss=0.05968, over 1513702.87 frames. ], batch size: 19, lr: 3.04e-03, grad_scale: 16.0 +2023-02-07 08:02:13,446 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194549.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:02:30,146 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2075, 2.5276, 2.8508, 1.5905, 3.1437, 1.6561, 1.5506, 2.2063], + device='cuda:1'), covar=tensor([0.0792, 0.0360, 0.0283, 0.0730, 0.0417, 0.0989, 0.0865, 0.0521], + device='cuda:1'), in_proj_covar=tensor([0.0459, 0.0401, 0.0355, 0.0450, 0.0387, 0.0537, 0.0398, 0.0431], + device='cuda:1'), out_proj_covar=tensor([1.2215e-04, 1.0461e-04, 9.3020e-05, 1.1807e-04, 1.0144e-04, 1.5059e-04, + 1.0673e-04, 1.1351e-04], device='cuda:1') +2023-02-07 08:02:42,128 INFO [train.py:901] (1/4) Epoch 25, batch 600, loss[loss=0.169, simple_loss=0.2479, pruned_loss=0.04503, over 7792.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.286, pruned_loss=0.06009, over 1544282.26 frames. ], batch size: 19, lr: 3.04e-03, grad_scale: 16.0 +2023-02-07 08:02:48,737 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.589e+02 2.361e+02 2.970e+02 3.663e+02 1.001e+03, threshold=5.941e+02, percent-clipped=3.0 +2023-02-07 08:03:01,145 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 08:03:01,345 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194617.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:03:10,844 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5628, 1.3370, 1.7069, 1.3069, 0.9123, 1.4390, 1.5161, 1.4510], + device='cuda:1'), covar=tensor([0.0561, 0.1276, 0.1626, 0.1485, 0.0595, 0.1474, 0.0710, 0.0657], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0189, 0.0160, 0.0101, 0.0163, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 08:03:16,780 INFO [train.py:901] (1/4) Epoch 25, batch 650, loss[loss=0.2105, simple_loss=0.2928, pruned_loss=0.06412, over 7919.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.285, pruned_loss=0.05968, over 1556266.02 frames. ], batch size: 20, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:03:18,067 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194642.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:03:23,709 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0881, 1.9083, 2.3921, 2.1124, 2.4008, 2.0901, 1.9713, 1.6252], + device='cuda:1'), covar=tensor([0.4073, 0.3868, 0.1652, 0.2891, 0.1843, 0.2670, 0.1457, 0.3709], + device='cuda:1'), in_proj_covar=tensor([0.0942, 0.0995, 0.0814, 0.0961, 0.1001, 0.0904, 0.0754, 0.0830], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 08:03:25,636 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194652.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:03:45,809 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194680.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:03:52,562 INFO [train.py:901] (1/4) Epoch 25, batch 700, loss[loss=0.1724, simple_loss=0.2555, pruned_loss=0.04464, over 7544.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2852, pruned_loss=0.06014, over 1567943.10 frames. ], batch size: 18, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:04:00,047 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.448e+02 2.849e+02 3.638e+02 5.412e+02, threshold=5.698e+02, percent-clipped=0.0 +2023-02-07 08:04:09,640 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194715.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:04:16,394 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194724.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:04:27,404 INFO [train.py:901] (1/4) Epoch 25, batch 750, loss[loss=0.1904, simple_loss=0.2737, pruned_loss=0.05357, over 8195.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2844, pruned_loss=0.06004, over 1576864.61 frames. ], batch size: 23, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:04:49,463 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 08:04:58,546 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 08:05:03,382 INFO [train.py:901] (1/4) Epoch 25, batch 800, loss[loss=0.176, simple_loss=0.253, pruned_loss=0.04947, over 7248.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2841, pruned_loss=0.05962, over 1587883.37 frames. ], batch size: 16, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:05:07,603 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194795.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:05:11,589 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.325e+02 2.417e+02 2.991e+02 3.771e+02 6.788e+02, threshold=5.982e+02, percent-clipped=2.0 +2023-02-07 08:05:14,533 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194805.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:05:31,778 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194830.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:05:31,810 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194830.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:05:37,855 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194839.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:05:38,299 INFO [train.py:901] (1/4) Epoch 25, batch 850, loss[loss=0.2697, simple_loss=0.3298, pruned_loss=0.1048, over 8437.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2849, pruned_loss=0.05993, over 1593851.25 frames. ], batch size: 27, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:05:49,038 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3966, 2.3002, 1.8467, 2.1576, 2.0192, 1.5056, 1.8994, 1.8353], + device='cuda:1'), covar=tensor([0.1506, 0.0464, 0.1210, 0.0568, 0.0781, 0.1686, 0.1071, 0.1055], + device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0234, 0.0335, 0.0307, 0.0298, 0.0339, 0.0344, 0.0318], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 08:05:56,595 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:06:14,007 INFO [train.py:901] (1/4) Epoch 25, batch 900, loss[loss=0.1754, simple_loss=0.26, pruned_loss=0.04539, over 7980.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2837, pruned_loss=0.05943, over 1596840.54 frames. ], batch size: 21, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:06:22,178 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.491e+02 2.923e+02 3.701e+02 8.623e+02, threshold=5.846e+02, percent-clipped=3.0 +2023-02-07 08:06:49,617 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-07 08:06:49,722 INFO [train.py:901] (1/4) Epoch 25, batch 950, loss[loss=0.238, simple_loss=0.3209, pruned_loss=0.07754, over 8189.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2832, pruned_loss=0.05859, over 1601626.84 frames. ], batch size: 23, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:07:02,579 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0905, 1.6531, 1.8477, 1.6241, 1.1136, 1.6817, 1.9654, 1.8278], + device='cuda:1'), covar=tensor([0.0594, 0.1177, 0.1625, 0.1357, 0.0649, 0.1375, 0.0701, 0.0605], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0191, 0.0160, 0.0101, 0.0164, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 08:07:19,512 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 08:07:24,289 INFO [train.py:901] (1/4) Epoch 25, batch 1000, loss[loss=0.1989, simple_loss=0.2815, pruned_loss=0.05821, over 7934.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2838, pruned_loss=0.05889, over 1606480.57 frames. ], batch size: 20, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:07:29,065 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194996.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:07:32,163 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.541e+02 2.651e+02 3.101e+02 3.894e+02 6.477e+02, threshold=6.202e+02, percent-clipped=4.0 +2023-02-07 08:07:51,784 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195029.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:07:54,457 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 08:07:59,735 INFO [train.py:901] (1/4) Epoch 25, batch 1050, loss[loss=0.2149, simple_loss=0.2994, pruned_loss=0.06517, over 8493.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2848, pruned_loss=0.05928, over 1611746.58 frames. ], batch size: 49, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:08:06,381 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 08:08:07,162 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195051.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:14,480 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195062.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:08:23,829 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195076.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:23,884 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4827, 1.3076, 1.5591, 1.4083, 1.4659, 1.4764, 1.3364, 0.7460], + device='cuda:1'), covar=tensor([0.3653, 0.3149, 0.1498, 0.2202, 0.1682, 0.2248, 0.1387, 0.3456], + device='cuda:1'), in_proj_covar=tensor([0.0950, 0.1003, 0.0824, 0.0968, 0.1009, 0.0914, 0.0761, 0.0836], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 08:08:31,305 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195086.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:33,294 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195089.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:33,849 INFO [train.py:901] (1/4) Epoch 25, batch 1100, loss[loss=0.1711, simple_loss=0.2524, pruned_loss=0.04494, over 7937.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2849, pruned_loss=0.0598, over 1607521.93 frames. ], batch size: 20, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:08:36,607 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8381, 1.6900, 2.5892, 1.5251, 2.1348, 2.8771, 2.8636, 2.5913], + device='cuda:1'), covar=tensor([0.1024, 0.1450, 0.0688, 0.1882, 0.1885, 0.0300, 0.0726, 0.0454], + device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0324, 0.0288, 0.0317, 0.0316, 0.0275, 0.0430, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 08:08:37,355 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195095.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:08:41,229 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.797e+02 2.550e+02 3.152e+02 4.111e+02 6.650e+02, threshold=6.304e+02, percent-clipped=3.0 +2023-02-07 08:08:48,154 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195111.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:48,177 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195111.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:53,476 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8411, 5.9161, 5.1817, 2.7069, 5.1796, 5.6158, 5.4262, 5.4014], + device='cuda:1'), covar=tensor([0.0505, 0.0375, 0.0910, 0.4002, 0.0717, 0.0715, 0.1094, 0.0518], + device='cuda:1'), in_proj_covar=tensor([0.0531, 0.0450, 0.0432, 0.0545, 0.0434, 0.0452, 0.0426, 0.0396], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:08:54,921 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195120.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:09:09,218 INFO [train.py:901] (1/4) Epoch 25, batch 1150, loss[loss=0.2069, simple_loss=0.2817, pruned_loss=0.06606, over 7936.00 frames. ], tot_loss[loss=0.201, simple_loss=0.284, pruned_loss=0.059, over 1609899.39 frames. ], batch size: 20, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:09:16,846 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 08:09:43,684 INFO [train.py:901] (1/4) Epoch 25, batch 1200, loss[loss=0.2019, simple_loss=0.2819, pruned_loss=0.06098, over 8499.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2852, pruned_loss=0.05897, over 1613459.72 frames. ], batch size: 49, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:09:44,707 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 +2023-02-07 08:09:51,813 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.229e+02 2.843e+02 3.492e+02 1.399e+03, threshold=5.685e+02, percent-clipped=2.0 +2023-02-07 08:09:57,123 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195209.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:10:18,374 INFO [train.py:901] (1/4) Epoch 25, batch 1250, loss[loss=0.1865, simple_loss=0.2653, pruned_loss=0.05379, over 8060.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2849, pruned_loss=0.05919, over 1609536.79 frames. ], batch size: 20, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:10:34,136 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.67 vs. limit=5.0 +2023-02-07 08:10:53,103 INFO [train.py:901] (1/4) Epoch 25, batch 1300, loss[loss=0.1646, simple_loss=0.2577, pruned_loss=0.0358, over 7974.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2849, pruned_loss=0.05896, over 1609377.39 frames. ], batch size: 21, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:11:00,272 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.452e+02 2.850e+02 4.025e+02 1.071e+03, threshold=5.700e+02, percent-clipped=7.0 +2023-02-07 08:11:16,286 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195324.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:11:26,830 INFO [train.py:901] (1/4) Epoch 25, batch 1350, loss[loss=0.2184, simple_loss=0.2883, pruned_loss=0.07425, over 6831.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2852, pruned_loss=0.05935, over 1607827.50 frames. ], batch size: 15, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:11:45,672 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195367.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:11:50,112 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195373.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:11:57,386 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-07 08:12:02,401 INFO [train.py:901] (1/4) Epoch 25, batch 1400, loss[loss=0.1909, simple_loss=0.2522, pruned_loss=0.06477, over 7424.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2855, pruned_loss=0.05942, over 1612398.62 frames. ], batch size: 17, lr: 3.04e-03, grad_scale: 4.0 +2023-02-07 08:12:04,000 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195392.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:12:10,473 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.570e+02 2.915e+02 3.833e+02 8.465e+02, threshold=5.831e+02, percent-clipped=6.0 +2023-02-07 08:12:13,226 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195406.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:12:15,768 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195410.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:12:31,396 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195433.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:12:36,127 INFO [train.py:901] (1/4) Epoch 25, batch 1450, loss[loss=0.2255, simple_loss=0.3033, pruned_loss=0.07385, over 8498.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2861, pruned_loss=0.06039, over 1614504.34 frames. ], batch size: 26, lr: 3.04e-03, grad_scale: 4.0 +2023-02-07 08:12:36,343 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5252, 1.4839, 1.8279, 1.2218, 1.2539, 1.8464, 0.2078, 1.2234], + device='cuda:1'), covar=tensor([0.1508, 0.1200, 0.0402, 0.0947, 0.2257, 0.0446, 0.1924, 0.1248], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0201, 0.0130, 0.0221, 0.0273, 0.0139, 0.0171, 0.0196], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 08:12:44,157 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 08:13:10,183 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195488.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:13:11,409 INFO [train.py:901] (1/4) Epoch 25, batch 1500, loss[loss=0.2093, simple_loss=0.2949, pruned_loss=0.06187, over 8267.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.285, pruned_loss=0.05998, over 1612335.07 frames. ], batch size: 24, lr: 3.04e-03, grad_scale: 4.0 +2023-02-07 08:13:19,806 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.401e+02 3.375e+02 4.255e+02 1.024e+03, threshold=6.749e+02, percent-clipped=12.0 +2023-02-07 08:13:33,927 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195521.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:13:46,249 INFO [train.py:901] (1/4) Epoch 25, batch 1550, loss[loss=0.2162, simple_loss=0.2872, pruned_loss=0.07258, over 7980.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2849, pruned_loss=0.05984, over 1613416.42 frames. ], batch size: 21, lr: 3.04e-03, grad_scale: 4.0 +2023-02-07 08:13:51,909 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195548.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:14:14,116 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195580.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:14:21,362 INFO [train.py:901] (1/4) Epoch 25, batch 1600, loss[loss=0.1964, simple_loss=0.2866, pruned_loss=0.05311, over 7810.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2836, pruned_loss=0.05922, over 1612253.42 frames. ], batch size: 20, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:14:29,486 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.317e+02 3.105e+02 3.813e+02 7.132e+02, threshold=6.211e+02, percent-clipped=3.0 +2023-02-07 08:14:32,329 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195605.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:14:37,017 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0990, 1.8350, 2.2729, 1.9170, 2.1658, 2.1932, 2.0183, 1.1030], + device='cuda:1'), covar=tensor([0.5653, 0.4793, 0.2120, 0.3982, 0.2896, 0.3265, 0.1894, 0.5436], + device='cuda:1'), in_proj_covar=tensor([0.0954, 0.1009, 0.0826, 0.0977, 0.1019, 0.0919, 0.0765, 0.0842], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 08:14:55,981 INFO [train.py:901] (1/4) Epoch 25, batch 1650, loss[loss=0.1959, simple_loss=0.2774, pruned_loss=0.05721, over 8292.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2837, pruned_loss=0.05908, over 1616631.68 frames. ], batch size: 23, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:15:29,744 INFO [train.py:901] (1/4) Epoch 25, batch 1700, loss[loss=0.2226, simple_loss=0.3051, pruned_loss=0.07001, over 8326.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2843, pruned_loss=0.05921, over 1616669.28 frames. ], batch size: 25, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:15:38,029 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.590e+02 3.116e+02 3.996e+02 7.880e+02, threshold=6.232e+02, percent-clipped=2.0 +2023-02-07 08:16:05,348 INFO [train.py:901] (1/4) Epoch 25, batch 1750, loss[loss=0.1641, simple_loss=0.2477, pruned_loss=0.04026, over 7256.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2847, pruned_loss=0.05936, over 1616516.83 frames. ], batch size: 16, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:16:06,322 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5673, 2.9117, 2.4481, 3.9827, 1.6051, 2.0561, 2.3217, 2.8120], + device='cuda:1'), covar=tensor([0.0637, 0.0697, 0.0760, 0.0243, 0.1074, 0.1235, 0.0972, 0.0796], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0196, 0.0245, 0.0213, 0.0205, 0.0245, 0.0249, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 08:16:09,114 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195744.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:16:15,708 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195754.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:16:17,280 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1074, 1.9738, 2.4318, 2.0261, 2.4916, 2.2358, 1.9902, 1.3577], + device='cuda:1'), covar=tensor([0.5551, 0.4881, 0.1958, 0.3931, 0.2401, 0.3144, 0.2036, 0.5287], + device='cuda:1'), in_proj_covar=tensor([0.0953, 0.1008, 0.0825, 0.0976, 0.1016, 0.0917, 0.0764, 0.0841], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 08:16:26,012 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195769.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:16:31,522 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195777.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:16:40,098 INFO [train.py:901] (1/4) Epoch 25, batch 1800, loss[loss=0.1879, simple_loss=0.2728, pruned_loss=0.05149, over 8309.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2851, pruned_loss=0.05956, over 1616044.90 frames. ], batch size: 25, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:16:48,985 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.456e+02 2.857e+02 3.484e+02 7.816e+02, threshold=5.715e+02, percent-clipped=1.0 +2023-02-07 08:16:49,207 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195802.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:16:50,558 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195804.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:17:07,910 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195829.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:17:15,239 INFO [train.py:901] (1/4) Epoch 25, batch 1850, loss[loss=0.243, simple_loss=0.3233, pruned_loss=0.08141, over 8332.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2847, pruned_loss=0.05934, over 1618632.51 frames. ], batch size: 25, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:17:36,404 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195869.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:17:50,148 INFO [train.py:901] (1/4) Epoch 25, batch 1900, loss[loss=0.1773, simple_loss=0.264, pruned_loss=0.04534, over 8632.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.284, pruned_loss=0.05938, over 1616060.02 frames. ], batch size: 49, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:17:58,368 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.686e+02 3.045e+02 3.689e+02 8.196e+02, threshold=6.090e+02, percent-clipped=3.0 +2023-02-07 08:18:24,464 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 08:18:25,126 INFO [train.py:901] (1/4) Epoch 25, batch 1950, loss[loss=0.1998, simple_loss=0.2855, pruned_loss=0.05708, over 8602.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2843, pruned_loss=0.05949, over 1612979.59 frames. ], batch size: 31, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:18:37,848 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 08:18:57,294 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 08:19:00,594 INFO [train.py:901] (1/4) Epoch 25, batch 2000, loss[loss=0.1631, simple_loss=0.2498, pruned_loss=0.03822, over 7227.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2844, pruned_loss=0.05951, over 1613604.71 frames. ], batch size: 16, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:19:09,746 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.344e+02 2.823e+02 3.287e+02 7.423e+02, threshold=5.646e+02, percent-clipped=4.0 +2023-02-07 08:19:11,917 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5962, 1.9756, 2.9137, 1.5077, 2.0471, 1.9412, 1.6673, 2.1553], + device='cuda:1'), covar=tensor([0.1956, 0.2538, 0.0829, 0.4534, 0.2094, 0.3205, 0.2440, 0.2311], + device='cuda:1'), in_proj_covar=tensor([0.0532, 0.0622, 0.0556, 0.0658, 0.0655, 0.0602, 0.0549, 0.0638], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:19:33,236 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.90 vs. limit=5.0 +2023-02-07 08:19:36,109 INFO [train.py:901] (1/4) Epoch 25, batch 2050, loss[loss=0.1935, simple_loss=0.2893, pruned_loss=0.04885, over 8472.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.284, pruned_loss=0.05877, over 1614589.78 frames. ], batch size: 29, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:20:00,445 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.97 vs. limit=5.0 +2023-02-07 08:20:11,114 INFO [train.py:901] (1/4) Epoch 25, batch 2100, loss[loss=0.1925, simple_loss=0.2697, pruned_loss=0.05762, over 8049.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2854, pruned_loss=0.06012, over 1615846.64 frames. ], batch size: 20, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:20:20,388 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.403e+02 2.946e+02 3.659e+02 8.101e+02, threshold=5.892e+02, percent-clipped=3.0 +2023-02-07 08:20:23,911 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5837, 2.7042, 2.4359, 4.0762, 1.5248, 2.0347, 2.4206, 2.8033], + device='cuda:1'), covar=tensor([0.0684, 0.0854, 0.0792, 0.0250, 0.1127, 0.1250, 0.0973, 0.0837], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0195, 0.0243, 0.0212, 0.0204, 0.0244, 0.0248, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 08:20:25,826 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3723, 1.3591, 4.7460, 1.9662, 3.8076, 3.7549, 4.2393, 4.2011], + device='cuda:1'), covar=tensor([0.1192, 0.7374, 0.0842, 0.5310, 0.1869, 0.1626, 0.1064, 0.0997], + device='cuda:1'), in_proj_covar=tensor([0.0649, 0.0655, 0.0712, 0.0645, 0.0724, 0.0617, 0.0621, 0.0695], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:20:35,907 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196125.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:20:46,054 INFO [train.py:901] (1/4) Epoch 25, batch 2150, loss[loss=0.1666, simple_loss=0.2442, pruned_loss=0.04447, over 7444.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2858, pruned_loss=0.0601, over 1617594.10 frames. ], batch size: 17, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:20:54,040 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196150.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:21:22,077 INFO [train.py:901] (1/4) Epoch 25, batch 2200, loss[loss=0.1828, simple_loss=0.276, pruned_loss=0.04486, over 8242.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2845, pruned_loss=0.05959, over 1615917.15 frames. ], batch size: 24, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:21:30,653 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.555e+02 3.213e+02 4.289e+02 6.887e+02, threshold=6.426e+02, percent-clipped=5.0 +2023-02-07 08:21:37,545 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.7137, 4.7466, 4.1774, 2.1715, 4.1630, 4.3145, 4.1023, 4.1350], + device='cuda:1'), covar=tensor([0.0742, 0.0503, 0.1084, 0.4720, 0.0931, 0.0999, 0.1456, 0.0712], + device='cuda:1'), in_proj_covar=tensor([0.0532, 0.0449, 0.0435, 0.0546, 0.0433, 0.0452, 0.0427, 0.0396], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:21:56,962 INFO [train.py:901] (1/4) Epoch 25, batch 2250, loss[loss=0.2229, simple_loss=0.3076, pruned_loss=0.06909, over 8315.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2836, pruned_loss=0.05983, over 1608526.79 frames. ], batch size: 25, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:22:07,573 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1950, 1.6977, 1.8287, 1.6877, 1.4086, 1.7131, 2.1036, 1.9765], + device='cuda:1'), covar=tensor([0.0571, 0.1375, 0.1965, 0.1523, 0.0710, 0.1684, 0.0849, 0.0597], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0152, 0.0190, 0.0160, 0.0100, 0.0164, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 08:22:12,518 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6993, 2.4323, 1.8227, 2.2587, 2.2051, 1.6795, 2.0775, 2.1466], + device='cuda:1'), covar=tensor([0.1363, 0.0441, 0.1179, 0.0620, 0.0698, 0.1526, 0.0980, 0.0937], + device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0238, 0.0341, 0.0313, 0.0303, 0.0345, 0.0350, 0.0323], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 08:22:32,058 INFO [train.py:901] (1/4) Epoch 25, batch 2300, loss[loss=0.1842, simple_loss=0.2749, pruned_loss=0.04677, over 8468.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.285, pruned_loss=0.06037, over 1608382.81 frames. ], batch size: 25, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:22:40,955 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.420e+02 2.794e+02 3.530e+02 9.865e+02, threshold=5.587e+02, percent-clipped=2.0 +2023-02-07 08:23:04,008 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0846, 2.2617, 1.8282, 2.9519, 1.3239, 1.6821, 2.0023, 2.2498], + device='cuda:1'), covar=tensor([0.0672, 0.0675, 0.0900, 0.0288, 0.1072, 0.1231, 0.0830, 0.0781], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0194, 0.0242, 0.0210, 0.0203, 0.0243, 0.0247, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 08:23:07,217 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196339.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:23:07,753 INFO [train.py:901] (1/4) Epoch 25, batch 2350, loss[loss=0.1822, simple_loss=0.2617, pruned_loss=0.05137, over 7265.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2852, pruned_loss=0.0601, over 1609789.14 frames. ], batch size: 16, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:23:34,292 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196378.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:23:42,086 INFO [train.py:901] (1/4) Epoch 25, batch 2400, loss[loss=0.1893, simple_loss=0.2668, pruned_loss=0.05593, over 8503.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2852, pruned_loss=0.0603, over 1607445.16 frames. ], batch size: 26, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:23:50,280 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.420e+02 2.902e+02 3.432e+02 7.434e+02, threshold=5.805e+02, percent-clipped=2.0 +2023-02-07 08:24:17,357 INFO [train.py:901] (1/4) Epoch 25, batch 2450, loss[loss=0.254, simple_loss=0.3329, pruned_loss=0.08752, over 8488.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2859, pruned_loss=0.06018, over 1609942.37 frames. ], batch size: 49, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:24:51,935 INFO [train.py:901] (1/4) Epoch 25, batch 2500, loss[loss=0.1962, simple_loss=0.2961, pruned_loss=0.04817, over 8108.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2868, pruned_loss=0.06018, over 1612358.58 frames. ], batch size: 23, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:25:00,802 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 2.398e+02 2.858e+02 3.242e+02 5.404e+02, threshold=5.717e+02, percent-clipped=0.0 +2023-02-07 08:25:18,290 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6491, 1.4408, 1.7525, 1.3728, 0.9987, 1.4779, 1.5042, 1.3681], + device='cuda:1'), covar=tensor([0.0535, 0.1170, 0.1613, 0.1468, 0.0566, 0.1392, 0.0680, 0.0691], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0161, 0.0101, 0.0164, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 08:25:26,994 INFO [train.py:901] (1/4) Epoch 25, batch 2550, loss[loss=0.1619, simple_loss=0.2443, pruned_loss=0.03976, over 5521.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2856, pruned_loss=0.05948, over 1605530.86 frames. ], batch size: 12, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:25:54,367 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7739, 1.6489, 2.0311, 1.8009, 2.0518, 1.8181, 1.6901, 1.2089], + device='cuda:1'), covar=tensor([0.4523, 0.3942, 0.1766, 0.2928, 0.1949, 0.2855, 0.1656, 0.4217], + device='cuda:1'), in_proj_covar=tensor([0.0952, 0.1004, 0.0824, 0.0974, 0.1014, 0.0915, 0.0761, 0.0837], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 08:26:02,159 INFO [train.py:901] (1/4) Epoch 25, batch 2600, loss[loss=0.1689, simple_loss=0.2532, pruned_loss=0.04226, over 8132.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2843, pruned_loss=0.05879, over 1607183.92 frames. ], batch size: 22, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:26:06,429 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196596.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:26:07,757 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8669, 3.8365, 3.4750, 1.8980, 3.3984, 3.4889, 3.4181, 3.3383], + device='cuda:1'), covar=tensor([0.0943, 0.0666, 0.1262, 0.4306, 0.1058, 0.1067, 0.1394, 0.0901], + device='cuda:1'), in_proj_covar=tensor([0.0530, 0.0447, 0.0433, 0.0542, 0.0434, 0.0450, 0.0424, 0.0396], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:26:10,247 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 2.366e+02 2.911e+02 3.287e+02 8.101e+02, threshold=5.822e+02, percent-clipped=1.0 +2023-02-07 08:26:13,504 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 +2023-02-07 08:26:37,071 INFO [train.py:901] (1/4) Epoch 25, batch 2650, loss[loss=0.1916, simple_loss=0.2722, pruned_loss=0.05545, over 7434.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2834, pruned_loss=0.05844, over 1607010.29 frames. ], batch size: 17, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:27:08,210 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=196683.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:27:12,709 INFO [train.py:901] (1/4) Epoch 25, batch 2700, loss[loss=0.2158, simple_loss=0.2974, pruned_loss=0.06707, over 8480.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2827, pruned_loss=0.05847, over 1603910.15 frames. ], batch size: 28, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:27:20,586 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.452e+02 2.909e+02 3.648e+02 8.771e+02, threshold=5.818e+02, percent-clipped=3.0 +2023-02-07 08:27:33,979 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=196722.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:27:46,998 INFO [train.py:901] (1/4) Epoch 25, batch 2750, loss[loss=0.1874, simple_loss=0.277, pruned_loss=0.04888, over 8460.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2828, pruned_loss=0.0588, over 1608135.04 frames. ], batch size: 25, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:27:55,331 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7655, 1.9087, 2.0406, 1.4427, 2.1095, 1.5243, 0.7799, 1.9297], + device='cuda:1'), covar=tensor([0.0640, 0.0419, 0.0318, 0.0639, 0.0571, 0.0924, 0.0986, 0.0350], + device='cuda:1'), in_proj_covar=tensor([0.0465, 0.0404, 0.0360, 0.0456, 0.0388, 0.0543, 0.0402, 0.0432], + device='cuda:1'), out_proj_covar=tensor([1.2356e-04, 1.0535e-04, 9.4351e-05, 1.1950e-04, 1.0162e-04, 1.5227e-04, + 1.0762e-04, 1.1362e-04], device='cuda:1') +2023-02-07 08:28:11,179 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.00 vs. limit=5.0 +2023-02-07 08:28:22,175 INFO [train.py:901] (1/4) Epoch 25, batch 2800, loss[loss=0.1914, simple_loss=0.2603, pruned_loss=0.06128, over 7531.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2837, pruned_loss=0.0589, over 1611947.63 frames. ], batch size: 18, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:28:27,858 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196797.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:28:28,543 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196798.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:28:31,145 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.445e+02 2.946e+02 3.604e+02 6.151e+02, threshold=5.892e+02, percent-clipped=2.0 +2023-02-07 08:28:52,423 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 08:28:54,961 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196837.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:28:56,679 INFO [train.py:901] (1/4) Epoch 25, batch 2850, loss[loss=0.2225, simple_loss=0.3005, pruned_loss=0.07229, over 7971.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2848, pruned_loss=0.05928, over 1611553.18 frames. ], batch size: 21, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:29:19,406 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196872.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:29:32,401 INFO [train.py:901] (1/4) Epoch 25, batch 2900, loss[loss=0.2076, simple_loss=0.3209, pruned_loss=0.04708, over 8236.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2839, pruned_loss=0.05844, over 1614065.02 frames. ], batch size: 24, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:29:39,447 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196899.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:29:41,313 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.504e+02 3.053e+02 3.742e+02 6.617e+02, threshold=6.106e+02, percent-clipped=2.0 +2023-02-07 08:30:08,127 INFO [train.py:901] (1/4) Epoch 25, batch 2950, loss[loss=0.1788, simple_loss=0.2587, pruned_loss=0.04942, over 8248.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.284, pruned_loss=0.0584, over 1617266.93 frames. ], batch size: 22, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:30:08,205 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=196940.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:30:08,829 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 08:30:19,843 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-07 08:30:41,608 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-07 08:30:42,486 INFO [train.py:901] (1/4) Epoch 25, batch 3000, loss[loss=0.2239, simple_loss=0.3176, pruned_loss=0.06512, over 8515.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2839, pruned_loss=0.05867, over 1620842.98 frames. ], batch size: 28, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:30:42,487 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 08:30:55,644 INFO [train.py:935] (1/4) Epoch 25, validation: loss=0.1722, simple_loss=0.2721, pruned_loss=0.03618, over 944034.00 frames. +2023-02-07 08:30:55,645 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 08:31:03,959 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 2.477e+02 2.955e+02 3.925e+02 7.788e+02, threshold=5.910e+02, percent-clipped=1.0 +2023-02-07 08:31:30,698 INFO [train.py:901] (1/4) Epoch 25, batch 3050, loss[loss=0.1858, simple_loss=0.2681, pruned_loss=0.05178, over 7652.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.283, pruned_loss=0.05894, over 1612988.81 frames. ], batch size: 19, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:31:40,519 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197054.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:31:41,157 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197055.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:31:57,862 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197079.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:32:05,195 INFO [train.py:901] (1/4) Epoch 25, batch 3100, loss[loss=0.2175, simple_loss=0.3008, pruned_loss=0.06711, over 8205.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.284, pruned_loss=0.05939, over 1611666.66 frames. ], batch size: 48, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:32:07,487 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197093.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:32:13,235 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.425e+02 3.089e+02 3.818e+02 7.102e+02, threshold=6.178e+02, percent-clipped=3.0 +2023-02-07 08:32:24,976 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:32:40,204 INFO [train.py:901] (1/4) Epoch 25, batch 3150, loss[loss=0.1707, simple_loss=0.2659, pruned_loss=0.03771, over 8036.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2841, pruned_loss=0.05899, over 1618549.62 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:32:40,984 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197141.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:33:15,297 INFO [train.py:901] (1/4) Epoch 25, batch 3200, loss[loss=0.193, simple_loss=0.2859, pruned_loss=0.04999, over 8536.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2835, pruned_loss=0.05848, over 1616454.23 frames. ], batch size: 31, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:33:23,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.400e+02 2.739e+02 3.315e+02 1.024e+03, threshold=5.479e+02, percent-clipped=5.0 +2023-02-07 08:33:33,159 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197216.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:33:50,303 INFO [train.py:901] (1/4) Epoch 25, batch 3250, loss[loss=0.2079, simple_loss=0.2901, pruned_loss=0.0628, over 8251.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2839, pruned_loss=0.05852, over 1614197.54 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:33:52,456 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197243.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:34:02,152 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197256.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:34:25,384 INFO [train.py:901] (1/4) Epoch 25, batch 3300, loss[loss=0.1562, simple_loss=0.2355, pruned_loss=0.03845, over 7793.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2841, pruned_loss=0.05915, over 1610571.29 frames. ], batch size: 19, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:34:34,249 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.560e+02 3.230e+02 4.212e+02 8.703e+02, threshold=6.460e+02, percent-clipped=10.0 +2023-02-07 08:34:40,505 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197311.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:34:46,281 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-07 08:34:54,226 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197331.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:34:57,732 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197336.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:35:00,243 INFO [train.py:901] (1/4) Epoch 25, batch 3350, loss[loss=0.2217, simple_loss=0.3077, pruned_loss=0.06788, over 8722.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2847, pruned_loss=0.05955, over 1610168.61 frames. ], batch size: 40, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:35:13,348 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197358.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:35:17,466 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8107, 1.7248, 2.4173, 1.5260, 1.2711, 2.4094, 0.3842, 1.4524], + device='cuda:1'), covar=tensor([0.1584, 0.1289, 0.0363, 0.1129, 0.2517, 0.0392, 0.2089, 0.1296], + device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0200, 0.0130, 0.0220, 0.0271, 0.0139, 0.0170, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 08:35:36,172 INFO [train.py:901] (1/4) Epoch 25, batch 3400, loss[loss=0.2005, simple_loss=0.2941, pruned_loss=0.05346, over 8516.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2859, pruned_loss=0.05989, over 1612495.14 frames. ], batch size: 28, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:35:39,767 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5196, 2.1636, 3.5579, 1.6029, 1.4334, 3.5915, 0.4560, 2.0473], + device='cuda:1'), covar=tensor([0.1458, 0.1258, 0.0207, 0.1932, 0.2751, 0.0200, 0.2168, 0.1279], + device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0199, 0.0130, 0.0220, 0.0271, 0.0138, 0.0170, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 08:35:44,270 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.489e+02 3.044e+02 3.734e+02 7.163e+02, threshold=6.087e+02, percent-clipped=2.0 +2023-02-07 08:36:06,317 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197433.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:36:11,060 INFO [train.py:901] (1/4) Epoch 25, batch 3450, loss[loss=0.1899, simple_loss=0.2787, pruned_loss=0.05052, over 8100.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2859, pruned_loss=0.05985, over 1613559.93 frames. ], batch size: 23, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:36:16,731 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3857, 2.1699, 1.7278, 2.0094, 1.7643, 1.5146, 1.7474, 1.7481], + device='cuda:1'), covar=tensor([0.1261, 0.0398, 0.1246, 0.0515, 0.0735, 0.1525, 0.0953, 0.0848], + device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0236, 0.0340, 0.0311, 0.0300, 0.0345, 0.0350, 0.0323], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 08:36:46,261 INFO [train.py:901] (1/4) Epoch 25, batch 3500, loss[loss=0.1907, simple_loss=0.2727, pruned_loss=0.05433, over 7809.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2846, pruned_loss=0.05912, over 1608531.21 frames. ], batch size: 20, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:36:54,913 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.387e+02 2.953e+02 3.537e+02 5.869e+02, threshold=5.907e+02, percent-clipped=0.0 +2023-02-07 08:37:02,040 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197512.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:07,264 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 08:37:17,733 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197534.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:19,765 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197537.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:21,600 INFO [train.py:901] (1/4) Epoch 25, batch 3550, loss[loss=0.1614, simple_loss=0.2389, pruned_loss=0.04193, over 7542.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2846, pruned_loss=0.05913, over 1608251.86 frames. ], batch size: 18, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:37:22,464 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.1969, 1.0281, 1.2968, 1.0403, 0.9231, 1.3313, 0.0485, 0.9502], + device='cuda:1'), covar=tensor([0.1391, 0.1263, 0.0465, 0.0673, 0.2503, 0.0511, 0.2019, 0.1120], + device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0201, 0.0131, 0.0221, 0.0273, 0.0139, 0.0172, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 08:37:24,422 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4068, 2.2329, 3.1946, 2.4796, 3.0973, 2.5083, 2.3088, 1.8497], + device='cuda:1'), covar=tensor([0.5716, 0.5361, 0.2081, 0.3956, 0.2417, 0.2980, 0.1878, 0.5529], + device='cuda:1'), in_proj_covar=tensor([0.0951, 0.1006, 0.0822, 0.0974, 0.1014, 0.0914, 0.0763, 0.0840], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 08:37:37,368 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197563.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:52,450 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3223, 2.5397, 2.0454, 2.9758, 1.5656, 1.9238, 2.2912, 2.4912], + device='cuda:1'), covar=tensor([0.0635, 0.0692, 0.0856, 0.0382, 0.0977, 0.1194, 0.0691, 0.0698], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0194, 0.0245, 0.0212, 0.0204, 0.0247, 0.0248, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 08:37:54,484 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197587.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:56,271 INFO [train.py:901] (1/4) Epoch 25, batch 3600, loss[loss=0.1814, simple_loss=0.2695, pruned_loss=0.04664, over 7932.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2839, pruned_loss=0.05857, over 1608834.85 frames. ], batch size: 20, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:38:05,202 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.311e+02 2.881e+02 3.803e+02 6.346e+02, threshold=5.762e+02, percent-clipped=1.0 +2023-02-07 08:38:12,112 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197612.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:38:13,511 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197614.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:38:31,215 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197639.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:38:31,672 INFO [train.py:901] (1/4) Epoch 25, batch 3650, loss[loss=0.2067, simple_loss=0.2843, pruned_loss=0.06456, over 8084.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2836, pruned_loss=0.05838, over 1611137.80 frames. ], batch size: 21, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:38:40,978 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7867, 2.3420, 3.6320, 1.8849, 1.9377, 3.6474, 0.6284, 2.1433], + device='cuda:1'), covar=tensor([0.1651, 0.1174, 0.0221, 0.1621, 0.2425, 0.0282, 0.2218, 0.1518], + device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0201, 0.0131, 0.0221, 0.0273, 0.0139, 0.0171, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 08:38:54,183 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4534, 1.7088, 2.1069, 1.3981, 1.5015, 1.7405, 1.5459, 1.5372], + device='cuda:1'), covar=tensor([0.1848, 0.2407, 0.1007, 0.4277, 0.2001, 0.3072, 0.2262, 0.2085], + device='cuda:1'), in_proj_covar=tensor([0.0532, 0.0621, 0.0557, 0.0659, 0.0654, 0.0603, 0.0549, 0.0637], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:39:06,721 INFO [train.py:901] (1/4) Epoch 25, batch 3700, loss[loss=0.1923, simple_loss=0.2825, pruned_loss=0.05104, over 8098.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2834, pruned_loss=0.05818, over 1610139.82 frames. ], batch size: 23, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:39:09,552 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 08:39:15,742 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.484e+02 2.942e+02 3.783e+02 7.174e+02, threshold=5.884e+02, percent-clipped=5.0 +2023-02-07 08:39:35,845 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1613, 1.5097, 4.3771, 1.6313, 3.9162, 3.6723, 3.9887, 3.8833], + device='cuda:1'), covar=tensor([0.0615, 0.4543, 0.0591, 0.4193, 0.1189, 0.0995, 0.0587, 0.0717], + device='cuda:1'), in_proj_covar=tensor([0.0655, 0.0660, 0.0721, 0.0650, 0.0730, 0.0624, 0.0626, 0.0701], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:39:40,724 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 08:39:43,105 INFO [train.py:901] (1/4) Epoch 25, batch 3750, loss[loss=0.2078, simple_loss=0.2884, pruned_loss=0.06359, over 8751.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2857, pruned_loss=0.05968, over 1616309.57 frames. ], batch size: 39, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:40:09,409 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197777.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:40:18,198 INFO [train.py:901] (1/4) Epoch 25, batch 3800, loss[loss=0.2119, simple_loss=0.2976, pruned_loss=0.06313, over 8540.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.285, pruned_loss=0.05986, over 1611188.71 frames. ], batch size: 49, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:40:26,490 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.549e+02 3.044e+02 3.681e+02 9.424e+02, threshold=6.087e+02, percent-clipped=5.0 +2023-02-07 08:40:34,308 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-07 08:40:43,132 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-02-07 08:40:53,469 INFO [train.py:901] (1/4) Epoch 25, batch 3850, loss[loss=0.2153, simple_loss=0.3044, pruned_loss=0.06309, over 8090.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2848, pruned_loss=0.05993, over 1611883.33 frames. ], batch size: 21, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:40:57,614 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9826, 2.0511, 1.7636, 2.5975, 1.2905, 1.6094, 1.9471, 2.0146], + device='cuda:1'), covar=tensor([0.0689, 0.0730, 0.0912, 0.0404, 0.0990, 0.1308, 0.0760, 0.0769], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0195, 0.0245, 0.0211, 0.0205, 0.0247, 0.0247, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 08:41:12,906 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 08:41:19,586 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197878.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:41:28,516 INFO [train.py:901] (1/4) Epoch 25, batch 3900, loss[loss=0.1739, simple_loss=0.257, pruned_loss=0.04538, over 7805.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.284, pruned_loss=0.05951, over 1609933.33 frames. ], batch size: 19, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:41:29,963 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197892.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:41:36,382 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.445e+02 2.982e+02 3.609e+02 8.629e+02, threshold=5.963e+02, percent-clipped=3.0 +2023-02-07 08:41:39,826 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197907.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:41:41,929 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197910.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:42:02,723 INFO [train.py:901] (1/4) Epoch 25, batch 3950, loss[loss=0.1838, simple_loss=0.2751, pruned_loss=0.04623, over 8135.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.284, pruned_loss=0.0597, over 1609379.52 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:42:24,241 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1652, 1.3056, 4.3414, 1.6603, 3.9085, 3.6089, 3.9580, 3.8471], + device='cuda:1'), covar=tensor([0.0596, 0.4843, 0.0539, 0.3939, 0.1078, 0.0923, 0.0568, 0.0664], + device='cuda:1'), in_proj_covar=tensor([0.0648, 0.0653, 0.0714, 0.0642, 0.0723, 0.0618, 0.0621, 0.0692], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:42:37,824 INFO [train.py:901] (1/4) Epoch 25, batch 4000, loss[loss=0.2177, simple_loss=0.298, pruned_loss=0.06866, over 8606.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2842, pruned_loss=0.05962, over 1611754.41 frames. ], batch size: 34, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:42:40,144 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197993.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:42:47,779 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.312e+02 2.768e+02 3.562e+02 7.475e+02, threshold=5.536e+02, percent-clipped=2.0 +2023-02-07 08:43:01,547 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198022.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:43:14,008 INFO [train.py:901] (1/4) Epoch 25, batch 4050, loss[loss=0.1906, simple_loss=0.2853, pruned_loss=0.04796, over 8318.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2854, pruned_loss=0.06006, over 1619197.68 frames. ], batch size: 25, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:43:26,785 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-07 08:43:41,815 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 08:43:48,793 INFO [train.py:901] (1/4) Epoch 25, batch 4100, loss[loss=0.2471, simple_loss=0.331, pruned_loss=0.08163, over 8474.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2855, pruned_loss=0.06018, over 1616271.71 frames. ], batch size: 29, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:43:55,128 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198099.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:43:57,001 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.413e+02 2.876e+02 3.434e+02 5.292e+02, threshold=5.752e+02, percent-clipped=1.0 +2023-02-07 08:44:24,280 INFO [train.py:901] (1/4) Epoch 25, batch 4150, loss[loss=0.2362, simple_loss=0.3099, pruned_loss=0.08127, over 8020.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2856, pruned_loss=0.05998, over 1615285.49 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:44:29,938 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198148.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:44:47,408 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198173.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:44:58,967 INFO [train.py:901] (1/4) Epoch 25, batch 4200, loss[loss=0.1735, simple_loss=0.251, pruned_loss=0.04797, over 7644.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2852, pruned_loss=0.05936, over 1618967.59 frames. ], batch size: 19, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:45:08,033 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.351e+02 3.091e+02 3.845e+02 7.201e+02, threshold=6.182e+02, percent-clipped=4.0 +2023-02-07 08:45:09,432 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 08:45:33,126 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 08:45:35,171 INFO [train.py:901] (1/4) Epoch 25, batch 4250, loss[loss=0.1876, simple_loss=0.2742, pruned_loss=0.05052, over 8251.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2845, pruned_loss=0.05968, over 1608646.26 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:45:40,414 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-07 08:45:41,594 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198249.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:45:44,713 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=198254.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:45:59,329 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198274.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:46:02,106 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198278.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:46:07,395 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5542, 1.6558, 2.0956, 1.3842, 1.5570, 1.7121, 1.6174, 1.4194], + device='cuda:1'), covar=tensor([0.2170, 0.2780, 0.1156, 0.5068, 0.2256, 0.3802, 0.2666, 0.2482], + device='cuda:1'), in_proj_covar=tensor([0.0531, 0.0621, 0.0555, 0.0658, 0.0653, 0.0604, 0.0550, 0.0636], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:46:09,926 INFO [train.py:901] (1/4) Epoch 25, batch 4300, loss[loss=0.1772, simple_loss=0.2725, pruned_loss=0.04091, over 8028.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2843, pruned_loss=0.05957, over 1611185.01 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:46:18,872 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.314e+02 2.735e+02 3.533e+02 6.805e+02, threshold=5.471e+02, percent-clipped=1.0 +2023-02-07 08:46:19,821 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198303.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:46:45,626 INFO [train.py:901] (1/4) Epoch 25, batch 4350, loss[loss=0.1705, simple_loss=0.2671, pruned_loss=0.03691, over 8458.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2834, pruned_loss=0.05892, over 1611031.08 frames. ], batch size: 25, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:47:04,272 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 08:47:06,479 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198369.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:47:08,431 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7353, 1.8396, 1.5590, 2.2667, 1.0826, 1.4421, 1.7497, 1.8447], + device='cuda:1'), covar=tensor([0.0811, 0.0793, 0.0976, 0.0445, 0.1065, 0.1382, 0.0786, 0.0799], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0194, 0.0245, 0.0212, 0.0204, 0.0247, 0.0248, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 08:47:21,573 INFO [train.py:901] (1/4) Epoch 25, batch 4400, loss[loss=0.2063, simple_loss=0.2858, pruned_loss=0.0634, over 8075.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2821, pruned_loss=0.0583, over 1610046.21 frames. ], batch size: 21, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:47:29,520 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.496e+02 2.496e+02 2.935e+02 3.768e+02 7.665e+02, threshold=5.870e+02, percent-clipped=6.0 +2023-02-07 08:47:45,265 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 08:47:56,744 INFO [train.py:901] (1/4) Epoch 25, batch 4450, loss[loss=0.2176, simple_loss=0.3076, pruned_loss=0.06377, over 8467.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2828, pruned_loss=0.05843, over 1611996.92 frames. ], batch size: 27, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:47:58,934 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=198443.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:48:04,421 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8436, 6.0370, 5.1317, 2.9567, 5.2842, 5.6360, 5.3567, 5.4120], + device='cuda:1'), covar=tensor([0.0462, 0.0308, 0.0795, 0.3787, 0.0705, 0.0711, 0.1042, 0.0586], + device='cuda:1'), in_proj_covar=tensor([0.0535, 0.0451, 0.0438, 0.0549, 0.0436, 0.0456, 0.0427, 0.0400], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:48:31,978 INFO [train.py:901] (1/4) Epoch 25, batch 4500, loss[loss=0.1991, simple_loss=0.2905, pruned_loss=0.05391, over 8644.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2812, pruned_loss=0.05785, over 1603463.46 frames. ], batch size: 39, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:48:40,442 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.409e+02 2.274e+02 2.771e+02 3.541e+02 5.802e+02, threshold=5.543e+02, percent-clipped=0.0 +2023-02-07 08:48:40,491 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 08:48:51,917 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:49:08,839 INFO [train.py:901] (1/4) Epoch 25, batch 4550, loss[loss=0.1843, simple_loss=0.2752, pruned_loss=0.0467, over 8484.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2819, pruned_loss=0.05784, over 1608432.66 frames. ], batch size: 29, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:49:14,807 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-07 08:49:22,073 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198558.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:49:44,786 INFO [train.py:901] (1/4) Epoch 25, batch 4600, loss[loss=0.1929, simple_loss=0.2763, pruned_loss=0.05475, over 8233.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2829, pruned_loss=0.05867, over 1607894.88 frames. ], batch size: 22, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:49:52,976 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.449e+02 2.940e+02 3.432e+02 8.422e+02, threshold=5.881e+02, percent-clipped=6.0 +2023-02-07 08:50:09,324 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198625.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:50:14,651 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198633.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:50:19,152 INFO [train.py:901] (1/4) Epoch 25, batch 4650, loss[loss=0.2183, simple_loss=0.3011, pruned_loss=0.06772, over 8576.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.284, pruned_loss=0.05962, over 1605848.34 frames. ], batch size: 39, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:50:26,821 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198650.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:50:54,510 INFO [train.py:901] (1/4) Epoch 25, batch 4700, loss[loss=0.1843, simple_loss=0.2779, pruned_loss=0.04538, over 8478.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2839, pruned_loss=0.05908, over 1609765.04 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:51:03,365 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.164e+02 2.735e+02 3.323e+02 7.623e+02, threshold=5.470e+02, percent-clipped=2.0 +2023-02-07 08:51:29,729 INFO [train.py:901] (1/4) Epoch 25, batch 4750, loss[loss=0.1596, simple_loss=0.2433, pruned_loss=0.03797, over 7987.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2839, pruned_loss=0.0591, over 1608668.90 frames. ], batch size: 21, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:51:38,026 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3775, 1.3980, 4.5942, 1.8187, 4.0674, 3.8075, 4.1477, 3.9908], + device='cuda:1'), covar=tensor([0.0568, 0.4758, 0.0514, 0.3831, 0.1102, 0.0940, 0.0579, 0.0721], + device='cuda:1'), in_proj_covar=tensor([0.0646, 0.0650, 0.0709, 0.0640, 0.0723, 0.0616, 0.0617, 0.0690], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:51:39,046 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-07 08:51:42,014 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 08:51:45,384 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 08:52:05,184 INFO [train.py:901] (1/4) Epoch 25, batch 4800, loss[loss=0.1844, simple_loss=0.2511, pruned_loss=0.05888, over 6781.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.282, pruned_loss=0.05819, over 1604221.87 frames. ], batch size: 15, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:52:13,388 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 2.392e+02 2.917e+02 3.409e+02 6.169e+02, threshold=5.835e+02, percent-clipped=3.0 +2023-02-07 08:52:22,061 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198814.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:52:36,134 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 08:52:39,627 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198839.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:52:40,098 INFO [train.py:901] (1/4) Epoch 25, batch 4850, loss[loss=0.2269, simple_loss=0.3093, pruned_loss=0.07223, over 8364.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2841, pruned_loss=0.05934, over 1612943.23 frames. ], batch size: 24, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:52:55,413 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=198861.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:53:01,716 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198870.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:53:15,867 INFO [train.py:901] (1/4) Epoch 25, batch 4900, loss[loss=0.2215, simple_loss=0.3001, pruned_loss=0.07146, over 8583.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2838, pruned_loss=0.05972, over 1611272.42 frames. ], batch size: 31, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:53:24,157 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198901.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:53:24,665 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 2.376e+02 2.954e+02 3.660e+02 6.336e+02, threshold=5.908e+02, percent-clipped=3.0 +2023-02-07 08:53:41,105 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-07 08:53:43,155 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 08:53:50,022 INFO [train.py:901] (1/4) Epoch 25, batch 4950, loss[loss=0.2359, simple_loss=0.3182, pruned_loss=0.07681, over 8522.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2842, pruned_loss=0.05935, over 1616945.65 frames. ], batch size: 39, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:53:54,413 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198945.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:54:15,937 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198976.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:54:16,534 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=198977.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:54:24,178 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1927, 1.9987, 2.6361, 2.1370, 2.5259, 2.2859, 2.0576, 1.4663], + device='cuda:1'), covar=tensor([0.5676, 0.5095, 0.2068, 0.3985, 0.2616, 0.3196, 0.2084, 0.5451], + device='cuda:1'), in_proj_covar=tensor([0.0949, 0.1002, 0.0817, 0.0969, 0.1011, 0.0914, 0.0758, 0.0835], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 08:54:25,251 INFO [train.py:901] (1/4) Epoch 25, batch 5000, loss[loss=0.2054, simple_loss=0.2959, pruned_loss=0.05747, over 8485.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2847, pruned_loss=0.05962, over 1614438.25 frames. ], batch size: 28, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:54:33,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.360e+02 2.883e+02 3.509e+02 6.136e+02, threshold=5.766e+02, percent-clipped=1.0 +2023-02-07 08:54:38,391 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.64 vs. limit=5.0 +2023-02-07 08:54:59,855 INFO [train.py:901] (1/4) Epoch 25, batch 5050, loss[loss=0.2149, simple_loss=0.3048, pruned_loss=0.0625, over 8452.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2846, pruned_loss=0.05976, over 1615778.84 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:55:14,363 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 08:55:35,776 INFO [train.py:901] (1/4) Epoch 25, batch 5100, loss[loss=0.1971, simple_loss=0.2934, pruned_loss=0.05038, over 8012.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.283, pruned_loss=0.05874, over 1617143.14 frames. ], batch size: 22, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:55:37,421 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199092.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:55:44,130 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.470e+02 3.005e+02 3.768e+02 7.063e+02, threshold=6.010e+02, percent-clipped=5.0 +2023-02-07 08:56:11,863 INFO [train.py:901] (1/4) Epoch 25, batch 5150, loss[loss=0.1938, simple_loss=0.293, pruned_loss=0.04735, over 8340.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.283, pruned_loss=0.05822, over 1618171.82 frames. ], batch size: 49, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:56:12,472 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-07 08:56:47,048 INFO [train.py:901] (1/4) Epoch 25, batch 5200, loss[loss=0.2609, simple_loss=0.3427, pruned_loss=0.08955, over 8325.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2836, pruned_loss=0.05828, over 1617886.68 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:56:49,910 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199194.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:56:55,026 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.800e+02 2.381e+02 2.894e+02 3.514e+02 1.206e+03, threshold=5.788e+02, percent-clipped=6.0 +2023-02-07 08:57:03,627 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 08:57:04,082 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:12,754 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 08:57:17,164 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199232.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:19,826 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0786, 1.6369, 1.6790, 1.4672, 0.9424, 1.5498, 1.7610, 1.5072], + device='cuda:1'), covar=tensor([0.0541, 0.1235, 0.1714, 0.1468, 0.0643, 0.1504, 0.0714, 0.0681], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0190, 0.0161, 0.0100, 0.0164, 0.0113, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 08:57:22,321 INFO [train.py:901] (1/4) Epoch 25, batch 5250, loss[loss=0.18, simple_loss=0.2579, pruned_loss=0.05106, over 7530.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2842, pruned_loss=0.05864, over 1620517.89 frames. ], batch size: 18, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:57:25,800 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199245.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:34,818 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199257.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:41,280 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2552, 3.0813, 2.8880, 1.6269, 2.8560, 2.9429, 2.7868, 2.8305], + device='cuda:1'), covar=tensor([0.1163, 0.0842, 0.1309, 0.4698, 0.1214, 0.1551, 0.1705, 0.1151], + device='cuda:1'), in_proj_covar=tensor([0.0542, 0.0456, 0.0442, 0.0555, 0.0439, 0.0460, 0.0433, 0.0404], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 08:57:44,764 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.7871, 2.1897, 3.4986, 1.8330, 1.7918, 3.5023, 0.5425, 2.1282], + device='cuda:1'), covar=tensor([0.1206, 0.1115, 0.0213, 0.1565, 0.2474, 0.0239, 0.2156, 0.1357], + device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0202, 0.0132, 0.0222, 0.0277, 0.0142, 0.0173, 0.0200], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 08:57:56,836 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199289.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:57,414 INFO [train.py:901] (1/4) Epoch 25, batch 5300, loss[loss=0.2122, simple_loss=0.3029, pruned_loss=0.06078, over 8326.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2833, pruned_loss=0.05786, over 1620413.97 frames. ], batch size: 26, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:58:05,706 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 2.313e+02 2.718e+02 3.488e+02 6.386e+02, threshold=5.437e+02, percent-clipped=3.0 +2023-02-07 08:58:25,244 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199329.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:58:32,743 INFO [train.py:901] (1/4) Epoch 25, batch 5350, loss[loss=0.2229, simple_loss=0.3059, pruned_loss=0.06999, over 8497.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2837, pruned_loss=0.05797, over 1623945.94 frames. ], batch size: 28, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:58:38,560 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199348.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:58:47,638 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199360.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:58:57,377 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199373.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:59:08,738 INFO [train.py:901] (1/4) Epoch 25, batch 5400, loss[loss=0.1806, simple_loss=0.2683, pruned_loss=0.04644, over 8539.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2838, pruned_loss=0.0581, over 1625146.03 frames. ], batch size: 28, lr: 3.01e-03, grad_scale: 32.0 +2023-02-07 08:59:18,156 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 2.292e+02 2.858e+02 3.757e+02 5.815e+02, threshold=5.716e+02, percent-clipped=3.0 +2023-02-07 08:59:18,354 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199404.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:59:22,502 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8423, 1.7511, 2.6913, 2.0823, 2.4527, 1.8983, 1.6673, 1.2400], + device='cuda:1'), covar=tensor([0.7496, 0.6200, 0.2193, 0.4158, 0.3075, 0.4340, 0.3007, 0.5711], + device='cuda:1'), in_proj_covar=tensor([0.0951, 0.1002, 0.0818, 0.0972, 0.1014, 0.0916, 0.0760, 0.0838], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 08:59:28,360 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199418.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:59:43,202 INFO [train.py:901] (1/4) Epoch 25, batch 5450, loss[loss=0.1732, simple_loss=0.2626, pruned_loss=0.04187, over 8246.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2837, pruned_loss=0.05827, over 1623328.99 frames. ], batch size: 22, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:00:08,099 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 09:00:08,226 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199476.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:00:17,974 INFO [train.py:901] (1/4) Epoch 25, batch 5500, loss[loss=0.2415, simple_loss=0.3032, pruned_loss=0.08992, over 7653.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2839, pruned_loss=0.0586, over 1624161.15 frames. ], batch size: 19, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:00:28,261 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.719e+02 2.278e+02 2.767e+02 3.622e+02 8.817e+02, threshold=5.534e+02, percent-clipped=3.0 +2023-02-07 09:00:33,821 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199512.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:00:52,113 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199538.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:00:53,324 INFO [train.py:901] (1/4) Epoch 25, batch 5550, loss[loss=0.2355, simple_loss=0.3088, pruned_loss=0.08112, over 8336.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2862, pruned_loss=0.05988, over 1625688.46 frames. ], batch size: 26, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:01:02,096 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199553.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:01:13,073 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.95 vs. limit=5.0 +2023-02-07 09:01:24,426 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199585.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:01:27,626 INFO [train.py:901] (1/4) Epoch 25, batch 5600, loss[loss=0.1968, simple_loss=0.281, pruned_loss=0.05631, over 8194.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2851, pruned_loss=0.05902, over 1622754.11 frames. ], batch size: 23, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:01:38,053 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.538e+02 3.116e+02 4.016e+02 1.228e+03, threshold=6.232e+02, percent-clipped=11.0 +2023-02-07 09:01:43,152 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199610.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:01:47,345 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199616.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:03,348 INFO [train.py:901] (1/4) Epoch 25, batch 5650, loss[loss=0.2673, simple_loss=0.3325, pruned_loss=0.101, over 7450.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2847, pruned_loss=0.05898, over 1617585.92 frames. ], batch size: 72, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:02:04,233 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199641.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:13,537 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199653.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:14,013 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 09:02:18,327 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199660.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:36,577 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199685.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:39,932 INFO [train.py:901] (1/4) Epoch 25, batch 5700, loss[loss=0.2053, simple_loss=0.2901, pruned_loss=0.06024, over 8353.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2841, pruned_loss=0.05889, over 1618628.83 frames. ], batch size: 24, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:02:49,771 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.200e+02 2.647e+02 3.419e+02 7.306e+02, threshold=5.294e+02, percent-clipped=3.0 +2023-02-07 09:03:12,405 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 09:03:16,064 INFO [train.py:901] (1/4) Epoch 25, batch 5750, loss[loss=0.184, simple_loss=0.2751, pruned_loss=0.04647, over 8032.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.283, pruned_loss=0.05834, over 1613421.88 frames. ], batch size: 22, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:03:16,266 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5709, 2.8816, 2.5183, 4.0827, 1.7096, 2.1550, 2.8696, 2.8556], + device='cuda:1'), covar=tensor([0.0732, 0.0719, 0.0743, 0.0234, 0.1143, 0.1233, 0.0825, 0.0852], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0193, 0.0243, 0.0209, 0.0204, 0.0245, 0.0247, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 09:03:21,540 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 09:03:30,823 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199762.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:03:39,188 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8336, 1.4044, 4.0322, 1.4474, 3.5263, 3.3118, 3.6470, 3.5118], + device='cuda:1'), covar=tensor([0.0705, 0.4598, 0.0584, 0.4014, 0.1329, 0.0964, 0.0651, 0.0817], + device='cuda:1'), in_proj_covar=tensor([0.0656, 0.0657, 0.0724, 0.0647, 0.0731, 0.0621, 0.0622, 0.0699], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:03:50,502 INFO [train.py:901] (1/4) Epoch 25, batch 5800, loss[loss=0.1998, simple_loss=0.284, pruned_loss=0.05782, over 8092.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2833, pruned_loss=0.05866, over 1611453.02 frames. ], batch size: 21, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:04:00,801 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.348e+02 2.869e+02 3.742e+02 6.332e+02, threshold=5.738e+02, percent-clipped=6.0 +2023-02-07 09:04:11,739 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199820.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:04:12,724 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 09:04:26,571 INFO [train.py:901] (1/4) Epoch 25, batch 5850, loss[loss=0.2572, simple_loss=0.3327, pruned_loss=0.09084, over 7150.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2841, pruned_loss=0.05921, over 1613373.61 frames. ], batch size: 71, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:04:37,361 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199856.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:04:51,703 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199877.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:05:00,999 INFO [train.py:901] (1/4) Epoch 25, batch 5900, loss[loss=0.2689, simple_loss=0.3314, pruned_loss=0.1032, over 7812.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2848, pruned_loss=0.05952, over 1615826.93 frames. ], batch size: 20, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:05:05,794 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199897.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:05:10,363 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.352e+02 2.828e+02 3.481e+02 7.421e+02, threshold=5.657e+02, percent-clipped=3.0 +2023-02-07 09:05:13,997 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199909.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:05:26,457 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199927.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:05:31,322 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199934.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:05:32,027 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199935.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:05:35,262 INFO [train.py:901] (1/4) Epoch 25, batch 5950, loss[loss=0.1955, simple_loss=0.2863, pruned_loss=0.05237, over 8570.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2864, pruned_loss=0.06049, over 1617294.01 frames. ], batch size: 31, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:05:54,662 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5157, 1.5295, 4.7721, 1.8565, 4.2652, 3.8971, 4.3159, 4.1498], + device='cuda:1'), covar=tensor([0.0554, 0.4642, 0.0445, 0.4158, 0.1100, 0.0949, 0.0548, 0.0665], + device='cuda:1'), in_proj_covar=tensor([0.0656, 0.0656, 0.0723, 0.0647, 0.0731, 0.0620, 0.0622, 0.0698], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:05:58,238 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199971.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:06:08,497 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3150, 2.0701, 2.7297, 2.2938, 2.7741, 2.3856, 2.2060, 1.5684], + device='cuda:1'), covar=tensor([0.5873, 0.5196, 0.2124, 0.4249, 0.2773, 0.3578, 0.2046, 0.6144], + device='cuda:1'), in_proj_covar=tensor([0.0955, 0.1004, 0.0822, 0.0976, 0.1017, 0.0917, 0.0763, 0.0839], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 09:06:11,076 INFO [train.py:901] (1/4) Epoch 25, batch 6000, loss[loss=0.2116, simple_loss=0.2868, pruned_loss=0.06822, over 8036.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2853, pruned_loss=0.05971, over 1610821.47 frames. ], batch size: 22, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:06:11,076 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 09:06:23,703 INFO [train.py:935] (1/4) Epoch 25, validation: loss=0.1725, simple_loss=0.2721, pruned_loss=0.03643, over 944034.00 frames. +2023-02-07 09:06:23,705 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 09:06:34,586 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.373e+02 2.952e+02 3.581e+02 7.260e+02, threshold=5.903e+02, percent-clipped=4.0 +2023-02-07 09:06:40,187 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200012.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:06:59,670 INFO [train.py:901] (1/4) Epoch 25, batch 6050, loss[loss=0.1695, simple_loss=0.2502, pruned_loss=0.04435, over 7984.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2851, pruned_loss=0.06021, over 1605554.03 frames. ], batch size: 21, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:07:35,033 INFO [train.py:901] (1/4) Epoch 25, batch 6100, loss[loss=0.1848, simple_loss=0.2709, pruned_loss=0.04938, over 8128.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2844, pruned_loss=0.05968, over 1606067.18 frames. ], batch size: 22, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:07:36,614 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6767, 1.5752, 4.9198, 1.8433, 4.3356, 4.0602, 4.4384, 4.3094], + device='cuda:1'), covar=tensor([0.0619, 0.4636, 0.0413, 0.4112, 0.1061, 0.0863, 0.0564, 0.0638], + device='cuda:1'), in_proj_covar=tensor([0.0657, 0.0656, 0.0724, 0.0647, 0.0732, 0.0621, 0.0624, 0.0699], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:07:38,670 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7182, 1.8235, 1.6084, 2.2619, 1.0815, 1.4281, 1.7064, 1.7870], + device='cuda:1'), covar=tensor([0.0732, 0.0741, 0.0877, 0.0469, 0.1069, 0.1340, 0.0755, 0.0741], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0194, 0.0245, 0.0211, 0.0205, 0.0247, 0.0249, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 09:07:45,359 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 2.345e+02 2.959e+02 3.596e+02 7.197e+02, threshold=5.919e+02, percent-clipped=3.0 +2023-02-07 09:07:54,353 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 09:08:05,861 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200133.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:08:11,247 INFO [train.py:901] (1/4) Epoch 25, batch 6150, loss[loss=0.2143, simple_loss=0.2864, pruned_loss=0.07106, over 8351.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2849, pruned_loss=0.06022, over 1610281.06 frames. ], batch size: 24, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:08:23,407 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200158.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:08:46,125 INFO [train.py:901] (1/4) Epoch 25, batch 6200, loss[loss=0.231, simple_loss=0.3178, pruned_loss=0.07207, over 8619.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2839, pruned_loss=0.05947, over 1612505.35 frames. ], batch size: 39, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:08:47,074 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200191.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:08:49,575 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200195.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:08:52,369 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6679, 1.5928, 5.8240, 2.3425, 5.2065, 4.8444, 5.4053, 5.2631], + device='cuda:1'), covar=tensor([0.0473, 0.5190, 0.0414, 0.4038, 0.1019, 0.0919, 0.0507, 0.0495], + device='cuda:1'), in_proj_covar=tensor([0.0659, 0.0656, 0.0725, 0.0649, 0.0733, 0.0623, 0.0625, 0.0701], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:08:55,700 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.314e+02 2.821e+02 3.535e+02 6.331e+02, threshold=5.643e+02, percent-clipped=2.0 +2023-02-07 09:09:04,326 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200216.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:09:13,093 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200227.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:09:21,657 INFO [train.py:901] (1/4) Epoch 25, batch 6250, loss[loss=0.1883, simple_loss=0.2638, pruned_loss=0.05643, over 7562.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2839, pruned_loss=0.05925, over 1612715.58 frames. ], batch size: 18, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:09:29,778 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200252.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:09:41,372 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200268.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:09:43,282 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=200271.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:09:46,148 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9646, 1.4577, 1.6957, 1.2967, 0.8994, 1.4306, 1.6006, 1.5728], + device='cuda:1'), covar=tensor([0.0500, 0.1278, 0.1617, 0.1470, 0.0587, 0.1453, 0.0715, 0.0650], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0153, 0.0189, 0.0160, 0.0100, 0.0163, 0.0112, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 09:09:56,027 INFO [train.py:901] (1/4) Epoch 25, batch 6300, loss[loss=0.1636, simple_loss=0.2418, pruned_loss=0.04264, over 7243.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2838, pruned_loss=0.05896, over 1609785.41 frames. ], batch size: 16, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:09:58,167 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200293.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:10:00,759 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200297.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:10:05,879 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-07 09:10:06,124 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.537e+02 3.046e+02 4.211e+02 7.306e+02, threshold=6.092e+02, percent-clipped=6.0 +2023-02-07 09:10:18,421 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.0455, 1.3706, 3.2791, 1.2388, 2.6456, 2.5596, 3.0023, 3.0044], + device='cuda:1'), covar=tensor([0.1425, 0.5570, 0.1466, 0.5341, 0.2634, 0.2334, 0.1169, 0.1262], + device='cuda:1'), in_proj_covar=tensor([0.0659, 0.0655, 0.0723, 0.0647, 0.0732, 0.0623, 0.0624, 0.0700], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:10:31,232 INFO [train.py:901] (1/4) Epoch 25, batch 6350, loss[loss=0.199, simple_loss=0.2846, pruned_loss=0.0567, over 7810.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2833, pruned_loss=0.05878, over 1607395.75 frames. ], batch size: 20, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:10:40,293 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8124, 5.8710, 5.1531, 2.4938, 5.2622, 5.6203, 5.3603, 5.3362], + device='cuda:1'), covar=tensor([0.0522, 0.0315, 0.0795, 0.4074, 0.0615, 0.0599, 0.1086, 0.0613], + device='cuda:1'), in_proj_covar=tensor([0.0531, 0.0448, 0.0434, 0.0546, 0.0432, 0.0451, 0.0427, 0.0397], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:11:03,771 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200386.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:11:06,377 INFO [train.py:901] (1/4) Epoch 25, batch 6400, loss[loss=0.1981, simple_loss=0.2924, pruned_loss=0.05187, over 8185.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.284, pruned_loss=0.05884, over 1613835.09 frames. ], batch size: 23, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:11:15,860 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.247e+02 2.600e+02 3.696e+02 8.014e+02, threshold=5.200e+02, percent-clipped=2.0 +2023-02-07 09:11:40,859 INFO [train.py:901] (1/4) Epoch 25, batch 6450, loss[loss=0.2083, simple_loss=0.2922, pruned_loss=0.06224, over 8345.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2832, pruned_loss=0.05847, over 1614084.82 frames. ], batch size: 48, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:11:49,291 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 09:11:55,915 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8008, 1.4268, 3.1059, 1.5639, 2.2899, 3.3664, 3.5187, 2.8652], + device='cuda:1'), covar=tensor([0.1190, 0.1774, 0.0335, 0.1925, 0.0999, 0.0253, 0.0498, 0.0540], + device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0320, 0.0285, 0.0314, 0.0314, 0.0271, 0.0429, 0.0302], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 09:12:16,063 INFO [train.py:901] (1/4) Epoch 25, batch 6500, loss[loss=0.2029, simple_loss=0.28, pruned_loss=0.0629, over 8445.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2823, pruned_loss=0.05803, over 1614072.08 frames. ], batch size: 29, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:12:26,030 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.181e+02 2.613e+02 3.190e+02 4.719e+02, threshold=5.226e+02, percent-clipped=0.0 +2023-02-07 09:12:49,391 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=200539.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:12:49,923 INFO [train.py:901] (1/4) Epoch 25, batch 6550, loss[loss=0.227, simple_loss=0.2929, pruned_loss=0.08053, over 7262.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.283, pruned_loss=0.05823, over 1616147.69 frames. ], batch size: 16, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:13:08,861 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-07 09:13:09,862 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 09:13:10,555 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4067, 4.3146, 3.9358, 2.3876, 3.8835, 3.9803, 3.9609, 3.7634], + device='cuda:1'), covar=tensor([0.0754, 0.0585, 0.1064, 0.4141, 0.0869, 0.1008, 0.1266, 0.0775], + device='cuda:1'), in_proj_covar=tensor([0.0534, 0.0452, 0.0438, 0.0552, 0.0437, 0.0457, 0.0431, 0.0400], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:13:26,032 INFO [train.py:901] (1/4) Epoch 25, batch 6600, loss[loss=0.222, simple_loss=0.2991, pruned_loss=0.07249, over 8340.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2833, pruned_loss=0.05857, over 1614250.86 frames. ], batch size: 26, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:13:30,831 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 09:13:35,560 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.529e+02 2.321e+02 2.722e+02 3.541e+02 8.507e+02, threshold=5.445e+02, percent-clipped=6.0 +2023-02-07 09:14:00,778 INFO [train.py:901] (1/4) Epoch 25, batch 6650, loss[loss=0.2109, simple_loss=0.2776, pruned_loss=0.07212, over 7527.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2835, pruned_loss=0.05833, over 1616457.77 frames. ], batch size: 18, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:14:01,601 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=200641.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:14:02,410 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200642.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:14:10,431 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200654.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:14:16,372 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200663.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:14:19,886 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200667.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:14:35,348 INFO [train.py:901] (1/4) Epoch 25, batch 6700, loss[loss=0.2043, simple_loss=0.2898, pruned_loss=0.05936, over 8454.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2833, pruned_loss=0.05841, over 1610745.61 frames. ], batch size: 25, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:14:45,627 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 2.443e+02 2.859e+02 3.397e+02 5.440e+02, threshold=5.717e+02, percent-clipped=0.0 +2023-02-07 09:15:01,824 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.96 vs. limit=5.0 +2023-02-07 09:15:10,911 INFO [train.py:901] (1/4) Epoch 25, batch 6750, loss[loss=0.1613, simple_loss=0.2408, pruned_loss=0.04092, over 7718.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2821, pruned_loss=0.05787, over 1613177.50 frames. ], batch size: 18, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:15:22,780 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200756.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:15:24,218 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7176, 2.3870, 3.8892, 1.5489, 3.0013, 2.2821, 1.9594, 2.8086], + device='cuda:1'), covar=tensor([0.1989, 0.2700, 0.0940, 0.4786, 0.1894, 0.3246, 0.2351, 0.2564], + device='cuda:1'), in_proj_covar=tensor([0.0535, 0.0625, 0.0558, 0.0662, 0.0661, 0.0605, 0.0554, 0.0644], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:15:47,012 INFO [train.py:901] (1/4) Epoch 25, batch 6800, loss[loss=0.2153, simple_loss=0.3018, pruned_loss=0.06444, over 8192.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2815, pruned_loss=0.05766, over 1617859.50 frames. ], batch size: 23, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:15:51,879 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 09:15:56,785 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.322e+02 2.853e+02 3.502e+02 6.162e+02, threshold=5.706e+02, percent-clipped=1.0 +2023-02-07 09:16:03,490 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-07 09:16:21,861 INFO [train.py:901] (1/4) Epoch 25, batch 6850, loss[loss=0.1496, simple_loss=0.2251, pruned_loss=0.037, over 6405.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.283, pruned_loss=0.05811, over 1621380.50 frames. ], batch size: 14, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:16:40,952 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 09:16:56,586 INFO [train.py:901] (1/4) Epoch 25, batch 6900, loss[loss=0.1625, simple_loss=0.252, pruned_loss=0.03648, over 7199.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2831, pruned_loss=0.0585, over 1615118.24 frames. ], batch size: 16, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:17:02,707 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.3670, 4.1977, 3.8465, 2.9096, 3.7658, 3.9660, 3.9435, 3.7567], + device='cuda:1'), covar=tensor([0.0604, 0.0657, 0.0948, 0.3276, 0.0922, 0.1217, 0.1120, 0.0882], + device='cuda:1'), in_proj_covar=tensor([0.0533, 0.0452, 0.0439, 0.0551, 0.0437, 0.0457, 0.0432, 0.0400], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:17:06,816 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.244e+02 2.770e+02 3.533e+02 6.127e+02, threshold=5.541e+02, percent-clipped=2.0 +2023-02-07 09:17:11,243 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200910.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:17:20,893 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 09:17:28,199 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200935.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:17:31,360 INFO [train.py:901] (1/4) Epoch 25, batch 6950, loss[loss=0.2045, simple_loss=0.2958, pruned_loss=0.0566, over 8328.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2829, pruned_loss=0.05842, over 1611566.43 frames. ], batch size: 25, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:17:50,991 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 09:17:58,863 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5674, 1.4986, 1.8558, 1.2405, 1.2084, 1.8320, 0.2326, 1.2360], + device='cuda:1'), covar=tensor([0.1645, 0.1199, 0.0432, 0.0837, 0.2375, 0.0545, 0.1804, 0.1138], + device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0199, 0.0130, 0.0219, 0.0273, 0.0140, 0.0170, 0.0196], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 09:18:07,790 INFO [train.py:901] (1/4) Epoch 25, batch 7000, loss[loss=0.1363, simple_loss=0.2166, pruned_loss=0.02797, over 7401.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2831, pruned_loss=0.05862, over 1612734.53 frames. ], batch size: 17, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:18:17,567 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.593e+02 3.026e+02 3.851e+02 8.547e+02, threshold=6.052e+02, percent-clipped=7.0 +2023-02-07 09:18:19,774 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=201007.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:18:23,071 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2388, 4.1826, 3.8388, 1.9263, 3.7354, 3.7416, 3.7323, 3.5857], + device='cuda:1'), covar=tensor([0.0834, 0.0560, 0.1010, 0.4303, 0.0975, 0.1079, 0.1373, 0.0818], + device='cuda:1'), in_proj_covar=tensor([0.0534, 0.0452, 0.0438, 0.0551, 0.0438, 0.0457, 0.0431, 0.0399], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:18:23,197 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201012.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:18:40,486 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201037.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:18:42,299 INFO [train.py:901] (1/4) Epoch 25, batch 7050, loss[loss=0.1695, simple_loss=0.2466, pruned_loss=0.0462, over 7208.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2824, pruned_loss=0.05816, over 1608893.39 frames. ], batch size: 16, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:19:16,835 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201089.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:19:17,378 INFO [train.py:901] (1/4) Epoch 25, batch 7100, loss[loss=0.1976, simple_loss=0.2766, pruned_loss=0.05928, over 7930.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2821, pruned_loss=0.05792, over 1609185.39 frames. ], batch size: 20, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:19:26,866 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.246e+02 2.728e+02 3.277e+02 5.322e+02, threshold=5.456e+02, percent-clipped=0.0 +2023-02-07 09:19:40,020 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201122.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:19:52,255 INFO [train.py:901] (1/4) Epoch 25, batch 7150, loss[loss=0.151, simple_loss=0.236, pruned_loss=0.03301, over 7787.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.282, pruned_loss=0.0584, over 1609216.98 frames. ], batch size: 19, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:19:53,558 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.97 vs. limit=5.0 +2023-02-07 09:20:28,435 INFO [train.py:901] (1/4) Epoch 25, batch 7200, loss[loss=0.2344, simple_loss=0.3273, pruned_loss=0.07075, over 8296.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2837, pruned_loss=0.05925, over 1611748.55 frames. ], batch size: 23, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:20:35,003 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-07 09:20:38,231 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.475e+02 3.123e+02 4.294e+02 9.608e+02, threshold=6.246e+02, percent-clipped=8.0 +2023-02-07 09:20:56,017 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-07 09:21:03,482 INFO [train.py:901] (1/4) Epoch 25, batch 7250, loss[loss=0.163, simple_loss=0.2537, pruned_loss=0.03612, over 7808.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2827, pruned_loss=0.05847, over 1609957.35 frames. ], batch size: 20, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:21:25,202 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201271.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:21:37,433 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0616, 1.5038, 3.4161, 1.5379, 2.2810, 3.7515, 3.8839, 3.1526], + device='cuda:1'), covar=tensor([0.1134, 0.1936, 0.0367, 0.2128, 0.1193, 0.0221, 0.0500, 0.0567], + device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0325, 0.0290, 0.0319, 0.0318, 0.0276, 0.0436, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 09:21:37,959 INFO [train.py:901] (1/4) Epoch 25, batch 7300, loss[loss=0.1841, simple_loss=0.2774, pruned_loss=0.04541, over 8459.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.283, pruned_loss=0.05881, over 1604766.19 frames. ], batch size: 25, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:21:39,378 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201292.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:21:48,706 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.341e+02 2.809e+02 3.464e+02 9.506e+02, threshold=5.617e+02, percent-clipped=4.0 +2023-02-07 09:21:50,352 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8533, 1.6398, 1.9844, 1.7766, 1.9197, 1.9593, 1.7928, 0.8797], + device='cuda:1'), covar=tensor([0.5966, 0.4940, 0.2249, 0.3845, 0.2647, 0.3291, 0.2149, 0.5347], + device='cuda:1'), in_proj_covar=tensor([0.0957, 0.1008, 0.0826, 0.0979, 0.1021, 0.0918, 0.0767, 0.0844], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 09:22:13,162 INFO [train.py:901] (1/4) Epoch 25, batch 7350, loss[loss=0.226, simple_loss=0.3002, pruned_loss=0.07587, over 8098.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2834, pruned_loss=0.05882, over 1607459.53 frames. ], batch size: 23, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:22:27,893 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.98 vs. limit=5.0 +2023-02-07 09:22:39,014 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 09:22:40,519 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201378.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:22:48,512 INFO [train.py:901] (1/4) Epoch 25, batch 7400, loss[loss=0.1592, simple_loss=0.2327, pruned_loss=0.04288, over 7218.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2827, pruned_loss=0.05874, over 1607161.86 frames. ], batch size: 16, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:22:57,494 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201403.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:22:57,970 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.231e+02 2.880e+02 3.857e+02 7.685e+02, threshold=5.759e+02, percent-clipped=5.0 +2023-02-07 09:22:58,695 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 09:23:19,301 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=201433.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:23:23,998 INFO [train.py:901] (1/4) Epoch 25, batch 7450, loss[loss=0.2173, simple_loss=0.2933, pruned_loss=0.07067, over 8036.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2843, pruned_loss=0.05977, over 1605730.92 frames. ], batch size: 22, lr: 2.99e-03, grad_scale: 16.0 +2023-02-07 09:23:37,760 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 09:23:59,782 INFO [train.py:901] (1/4) Epoch 25, batch 7500, loss[loss=0.1604, simple_loss=0.2401, pruned_loss=0.0403, over 8087.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2845, pruned_loss=0.05974, over 1609228.77 frames. ], batch size: 21, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:24:09,798 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.281e+02 2.758e+02 3.564e+02 6.593e+02, threshold=5.515e+02, percent-clipped=6.0 +2023-02-07 09:24:12,066 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6756, 2.4733, 1.6816, 2.3375, 2.1304, 1.4399, 2.0997, 2.2921], + device='cuda:1'), covar=tensor([0.1565, 0.0462, 0.1578, 0.0635, 0.0810, 0.1895, 0.1114, 0.0984], + device='cuda:1'), in_proj_covar=tensor([0.0360, 0.0239, 0.0345, 0.0316, 0.0304, 0.0348, 0.0352, 0.0324], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 09:24:34,700 INFO [train.py:901] (1/4) Epoch 25, batch 7550, loss[loss=0.1992, simple_loss=0.2855, pruned_loss=0.05646, over 8450.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2845, pruned_loss=0.05969, over 1611860.70 frames. ], batch size: 29, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:24:40,276 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201548.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:25:09,425 INFO [train.py:901] (1/4) Epoch 25, batch 7600, loss[loss=0.1597, simple_loss=0.2349, pruned_loss=0.04227, over 7430.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2849, pruned_loss=0.05964, over 1611069.36 frames. ], batch size: 17, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:25:20,530 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.467e+02 2.939e+02 3.909e+02 7.265e+02, threshold=5.878e+02, percent-clipped=5.0 +2023-02-07 09:25:27,306 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=201615.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:25:41,304 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=201636.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:25:43,929 INFO [train.py:901] (1/4) Epoch 25, batch 7650, loss[loss=0.1868, simple_loss=0.2785, pruned_loss=0.04752, over 7986.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2846, pruned_loss=0.05956, over 1609846.48 frames. ], batch size: 21, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:26:00,385 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201662.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:26:12,135 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201679.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:26:19,111 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.62 vs. limit=5.0 +2023-02-07 09:26:19,453 INFO [train.py:901] (1/4) Epoch 25, batch 7700, loss[loss=0.1946, simple_loss=0.2788, pruned_loss=0.05521, over 7813.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2844, pruned_loss=0.05934, over 1605418.05 frames. ], batch size: 20, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:26:20,539 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 09:26:20,995 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9064, 2.1157, 6.0372, 2.1583, 5.4270, 5.0833, 5.5562, 5.4706], + device='cuda:1'), covar=tensor([0.0433, 0.4467, 0.0392, 0.4202, 0.0959, 0.0772, 0.0498, 0.0486], + device='cuda:1'), in_proj_covar=tensor([0.0658, 0.0656, 0.0722, 0.0648, 0.0736, 0.0627, 0.0626, 0.0699], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:26:30,382 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.306e+02 2.805e+02 3.732e+02 7.115e+02, threshold=5.609e+02, percent-clipped=1.0 +2023-02-07 09:26:43,955 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201724.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:26:48,007 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201730.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:26:49,245 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 09:26:54,628 INFO [train.py:901] (1/4) Epoch 25, batch 7750, loss[loss=0.1916, simple_loss=0.284, pruned_loss=0.0496, over 8340.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2845, pruned_loss=0.05886, over 1610714.76 frames. ], batch size: 25, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:26:58,895 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6760, 1.9420, 2.0460, 1.4123, 2.0987, 1.5470, 0.5598, 1.9287], + device='cuda:1'), covar=tensor([0.0639, 0.0380, 0.0335, 0.0636, 0.0508, 0.0962, 0.1008, 0.0338], + device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0402, 0.0360, 0.0458, 0.0391, 0.0543, 0.0402, 0.0431], + device='cuda:1'), out_proj_covar=tensor([1.2294e-04, 1.0476e-04, 9.3949e-05, 1.1998e-04, 1.0235e-04, 1.5191e-04, + 1.0748e-04, 1.1315e-04], device='cuda:1') +2023-02-07 09:27:02,274 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201751.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:27:10,572 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2332, 1.0936, 1.3186, 1.0434, 1.0085, 1.3433, 0.0619, 0.9318], + device='cuda:1'), covar=tensor([0.1434, 0.1261, 0.0471, 0.0687, 0.2300, 0.0489, 0.1923, 0.1109], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0201, 0.0130, 0.0221, 0.0274, 0.0141, 0.0171, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 09:27:29,977 INFO [train.py:901] (1/4) Epoch 25, batch 7800, loss[loss=0.244, simple_loss=0.322, pruned_loss=0.083, over 8462.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2837, pruned_loss=0.05878, over 1610827.62 frames. ], batch size: 25, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:27:40,030 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201804.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:27:40,475 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.301e+02 2.955e+02 3.831e+02 1.047e+03, threshold=5.910e+02, percent-clipped=5.0 +2023-02-07 09:27:40,890 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-07 09:27:56,321 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201829.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:28:03,032 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.26 vs. limit=5.0 +2023-02-07 09:28:03,377 INFO [train.py:901] (1/4) Epoch 25, batch 7850, loss[loss=0.2112, simple_loss=0.3044, pruned_loss=0.05901, over 8481.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2835, pruned_loss=0.05847, over 1614370.95 frames. ], batch size: 28, lr: 2.99e-03, grad_scale: 4.0 +2023-02-07 09:28:23,070 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1698, 1.2809, 4.3937, 2.0470, 2.6047, 5.0509, 5.1042, 4.3730], + device='cuda:1'), covar=tensor([0.1218, 0.2116, 0.0254, 0.1871, 0.1178, 0.0149, 0.0337, 0.0534], + device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0322, 0.0289, 0.0318, 0.0316, 0.0273, 0.0433, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 09:28:36,527 INFO [train.py:901] (1/4) Epoch 25, batch 7900, loss[loss=0.1982, simple_loss=0.2862, pruned_loss=0.05506, over 8332.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2844, pruned_loss=0.05981, over 1611839.08 frames. ], batch size: 26, lr: 2.99e-03, grad_scale: 4.0 +2023-02-07 09:28:47,146 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.505e+02 3.187e+02 3.787e+02 7.491e+02, threshold=6.375e+02, percent-clipped=2.0 +2023-02-07 09:29:09,582 INFO [train.py:901] (1/4) Epoch 25, batch 7950, loss[loss=0.1982, simple_loss=0.2886, pruned_loss=0.05385, over 8338.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2837, pruned_loss=0.05928, over 1611023.47 frames. ], batch size: 26, lr: 2.99e-03, grad_scale: 4.0 +2023-02-07 09:29:18,131 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 +2023-02-07 09:29:40,346 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201986.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:29:42,796 INFO [train.py:901] (1/4) Epoch 25, batch 8000, loss[loss=0.1933, simple_loss=0.2715, pruned_loss=0.05751, over 8444.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2852, pruned_loss=0.06014, over 1615474.36 frames. ], batch size: 27, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:29:54,397 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 2.298e+02 3.131e+02 3.789e+02 6.155e+02, threshold=6.263e+02, percent-clipped=0.0 +2023-02-07 09:29:54,488 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=202006.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:29:55,329 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202007.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:29:55,931 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202008.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:29:58,062 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202011.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:30:06,543 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=202023.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:30:12,411 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202032.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:30:17,702 INFO [train.py:901] (1/4) Epoch 25, batch 8050, loss[loss=0.2219, simple_loss=0.2798, pruned_loss=0.08202, over 7544.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2838, pruned_loss=0.06018, over 1599077.08 frames. ], batch size: 18, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:30:36,918 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=202068.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:30:50,422 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 09:30:55,056 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 +2023-02-07 09:30:55,320 INFO [train.py:901] (1/4) Epoch 26, batch 0, loss[loss=0.2045, simple_loss=0.2841, pruned_loss=0.06243, over 7654.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2841, pruned_loss=0.06243, over 7654.00 frames. ], batch size: 19, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:30:55,320 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 09:31:04,640 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.3147, 2.1390, 1.6702, 1.8825, 1.7846, 1.5152, 1.7373, 1.7196], + device='cuda:1'), covar=tensor([0.1459, 0.0479, 0.1267, 0.0587, 0.0762, 0.1664, 0.1024, 0.0989], + device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0239, 0.0341, 0.0314, 0.0303, 0.0346, 0.0350, 0.0322], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 09:31:06,904 INFO [train.py:935] (1/4) Epoch 26, validation: loss=0.1717, simple_loss=0.2716, pruned_loss=0.03591, over 944034.00 frames. +2023-02-07 09:31:06,905 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 09:31:21,607 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 09:31:29,809 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 2.411e+02 2.993e+02 3.956e+02 9.314e+02, threshold=5.987e+02, percent-clipped=4.0 +2023-02-07 09:31:40,822 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202121.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:31:41,314 INFO [train.py:901] (1/4) Epoch 26, batch 50, loss[loss=0.2006, simple_loss=0.283, pruned_loss=0.05903, over 8523.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2866, pruned_loss=0.05892, over 368267.22 frames. ], batch size: 39, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:31:52,551 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202138.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:31:55,752 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 09:32:15,980 INFO [train.py:901] (1/4) Epoch 26, batch 100, loss[loss=0.1755, simple_loss=0.2559, pruned_loss=0.0475, over 7804.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2856, pruned_loss=0.0587, over 648336.14 frames. ], batch size: 20, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:32:18,604 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 09:32:23,609 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202183.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:32:40,575 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.435e+02 2.962e+02 3.649e+02 8.375e+02, threshold=5.925e+02, percent-clipped=4.0 +2023-02-07 09:32:51,108 INFO [train.py:901] (1/4) Epoch 26, batch 150, loss[loss=0.1835, simple_loss=0.2761, pruned_loss=0.04545, over 7976.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2857, pruned_loss=0.06003, over 862609.09 frames. ], batch size: 21, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:33:22,417 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5666, 1.4164, 4.7685, 1.7512, 4.2076, 3.9720, 4.3628, 4.2282], + device='cuda:1'), covar=tensor([0.0594, 0.4963, 0.0518, 0.4247, 0.1123, 0.0988, 0.0492, 0.0667], + device='cuda:1'), in_proj_covar=tensor([0.0660, 0.0653, 0.0721, 0.0646, 0.0730, 0.0626, 0.0625, 0.0699], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:33:26,394 INFO [train.py:901] (1/4) Epoch 26, batch 200, loss[loss=0.1844, simple_loss=0.277, pruned_loss=0.04589, over 8028.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.285, pruned_loss=0.05921, over 1029933.95 frames. ], batch size: 22, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:33:49,941 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 2.408e+02 2.928e+02 3.669e+02 9.390e+02, threshold=5.857e+02, percent-clipped=3.0 +2023-02-07 09:34:01,574 INFO [train.py:901] (1/4) Epoch 26, batch 250, loss[loss=0.2276, simple_loss=0.3259, pruned_loss=0.06465, over 8317.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2864, pruned_loss=0.05919, over 1167651.31 frames. ], batch size: 25, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:34:09,714 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 09:34:11,248 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3024, 2.6015, 2.8328, 1.8088, 3.1840, 1.9003, 1.6001, 2.2519], + device='cuda:1'), covar=tensor([0.0891, 0.0461, 0.0405, 0.0852, 0.0628, 0.0928, 0.1027, 0.0617], + device='cuda:1'), in_proj_covar=tensor([0.0458, 0.0398, 0.0355, 0.0451, 0.0384, 0.0536, 0.0395, 0.0426], + device='cuda:1'), out_proj_covar=tensor([1.2172e-04, 1.0358e-04, 9.2619e-05, 1.1821e-04, 1.0042e-04, 1.5006e-04, + 1.0588e-04, 1.1196e-04], device='cuda:1') +2023-02-07 09:34:19,974 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 09:34:22,428 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 09:34:22,770 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=202352.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:34:30,088 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 +2023-02-07 09:34:36,400 INFO [train.py:901] (1/4) Epoch 26, batch 300, loss[loss=0.1835, simple_loss=0.2525, pruned_loss=0.0573, over 7434.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2862, pruned_loss=0.05936, over 1265457.84 frames. ], batch size: 17, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:34:37,952 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7306, 1.7142, 2.3162, 1.4212, 1.3332, 2.3016, 0.3612, 1.4003], + device='cuda:1'), covar=tensor([0.1730, 0.1086, 0.0316, 0.1049, 0.2282, 0.0338, 0.1868, 0.1365], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0202, 0.0131, 0.0222, 0.0276, 0.0142, 0.0171, 0.0199], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 09:34:40,015 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202377.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:34:52,240 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202394.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:34:57,262 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202402.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:34:59,609 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 2.510e+02 3.033e+02 3.572e+02 1.183e+03, threshold=6.066e+02, percent-clipped=2.0 +2023-02-07 09:35:08,718 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202419.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:35:10,549 INFO [train.py:901] (1/4) Epoch 26, batch 350, loss[loss=0.1873, simple_loss=0.2792, pruned_loss=0.04771, over 8195.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2855, pruned_loss=0.05868, over 1348003.68 frames. ], batch size: 23, lr: 2.93e-03, grad_scale: 4.0 +2023-02-07 09:35:23,411 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202439.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:35:41,018 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202464.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:35:43,113 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202467.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:35:46,431 INFO [train.py:901] (1/4) Epoch 26, batch 400, loss[loss=0.2189, simple_loss=0.297, pruned_loss=0.07035, over 8470.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2845, pruned_loss=0.05796, over 1409459.51 frames. ], batch size: 25, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:36:11,056 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 2.504e+02 3.071e+02 3.633e+02 8.131e+02, threshold=6.142e+02, percent-clipped=3.0 +2023-02-07 09:36:21,078 INFO [train.py:901] (1/4) Epoch 26, batch 450, loss[loss=0.2053, simple_loss=0.2982, pruned_loss=0.05624, over 8297.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2842, pruned_loss=0.05826, over 1453502.52 frames. ], batch size: 23, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:36:55,499 INFO [train.py:901] (1/4) Epoch 26, batch 500, loss[loss=0.2077, simple_loss=0.2942, pruned_loss=0.06057, over 8505.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2852, pruned_loss=0.0593, over 1484300.36 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:37:11,283 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8522, 1.8398, 2.0384, 1.7157, 0.8972, 1.7866, 2.3283, 2.3793], + device='cuda:1'), covar=tensor([0.0458, 0.1134, 0.1543, 0.1332, 0.0610, 0.1356, 0.0577, 0.0540], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0190, 0.0160, 0.0100, 0.0163, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 09:37:19,246 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.396e+02 2.962e+02 4.085e+02 8.069e+02, threshold=5.924e+02, percent-clipped=6.0 +2023-02-07 09:37:29,369 INFO [train.py:901] (1/4) Epoch 26, batch 550, loss[loss=0.1953, simple_loss=0.2723, pruned_loss=0.05914, over 8130.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2848, pruned_loss=0.05944, over 1511657.31 frames. ], batch size: 22, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:37:45,557 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-07 09:38:00,898 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6077, 2.0502, 3.2128, 1.3696, 2.4157, 2.0702, 1.6946, 2.4922], + device='cuda:1'), covar=tensor([0.1937, 0.2575, 0.0869, 0.4832, 0.1945, 0.3231, 0.2444, 0.2257], + device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0629, 0.0559, 0.0663, 0.0661, 0.0609, 0.0557, 0.0645], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:38:05,142 INFO [train.py:901] (1/4) Epoch 26, batch 600, loss[loss=0.1754, simple_loss=0.2448, pruned_loss=0.05299, over 6815.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2839, pruned_loss=0.05892, over 1532296.43 frames. ], batch size: 15, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:38:21,711 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 09:38:29,023 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.445e+02 2.916e+02 3.512e+02 6.749e+02, threshold=5.833e+02, percent-clipped=3.0 +2023-02-07 09:38:38,959 INFO [train.py:901] (1/4) Epoch 26, batch 650, loss[loss=0.1778, simple_loss=0.2571, pruned_loss=0.04928, over 7929.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.284, pruned_loss=0.05889, over 1556516.52 frames. ], batch size: 20, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:38:39,868 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:38:57,385 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202748.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:39:14,894 INFO [train.py:901] (1/4) Epoch 26, batch 700, loss[loss=0.1878, simple_loss=0.283, pruned_loss=0.04627, over 8509.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2843, pruned_loss=0.0593, over 1564352.27 frames. ], batch size: 28, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:39:38,615 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.498e+02 3.029e+02 3.750e+02 8.351e+02, threshold=6.058e+02, percent-clipped=3.0 +2023-02-07 09:39:49,871 INFO [train.py:901] (1/4) Epoch 26, batch 750, loss[loss=0.2287, simple_loss=0.3079, pruned_loss=0.07475, over 8340.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2844, pruned_loss=0.05893, over 1578024.76 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:40:05,060 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 09:40:08,014 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202848.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:40:13,840 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 09:40:24,780 INFO [train.py:901] (1/4) Epoch 26, batch 800, loss[loss=0.2084, simple_loss=0.2956, pruned_loss=0.06063, over 8340.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2842, pruned_loss=0.05992, over 1580659.07 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:40:49,964 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.489e+02 2.818e+02 3.827e+02 7.280e+02, threshold=5.635e+02, percent-clipped=3.0 +2023-02-07 09:40:59,907 INFO [train.py:901] (1/4) Epoch 26, batch 850, loss[loss=0.1801, simple_loss=0.2713, pruned_loss=0.04448, over 8327.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2838, pruned_loss=0.0596, over 1589434.27 frames. ], batch size: 25, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:41:33,338 INFO [train.py:901] (1/4) Epoch 26, batch 900, loss[loss=0.2033, simple_loss=0.2979, pruned_loss=0.05437, over 8434.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2837, pruned_loss=0.05944, over 1592447.85 frames. ], batch size: 27, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:41:58,968 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.742e+02 3.265e+02 4.005e+02 6.934e+02, threshold=6.531e+02, percent-clipped=5.0 +2023-02-07 09:42:08,851 INFO [train.py:901] (1/4) Epoch 26, batch 950, loss[loss=0.2116, simple_loss=0.296, pruned_loss=0.0636, over 8494.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2845, pruned_loss=0.0601, over 1595470.61 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:42:16,930 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.1548, 3.0940, 2.8700, 1.9584, 2.8184, 2.7991, 2.8166, 2.6952], + device='cuda:1'), covar=tensor([0.0862, 0.0772, 0.1043, 0.3547, 0.0884, 0.1279, 0.1324, 0.0990], + device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0449, 0.0437, 0.0546, 0.0434, 0.0452, 0.0427, 0.0394], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:42:31,713 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 09:42:32,526 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203056.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:42:43,150 INFO [train.py:901] (1/4) Epoch 26, batch 1000, loss[loss=0.2066, simple_loss=0.2783, pruned_loss=0.06747, over 7254.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2852, pruned_loss=0.06064, over 1593509.68 frames. ], batch size: 16, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:42:57,331 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9039, 2.0522, 1.8657, 2.6361, 1.2514, 1.6516, 1.9582, 2.1030], + device='cuda:1'), covar=tensor([0.0765, 0.0845, 0.0936, 0.0371, 0.1086, 0.1309, 0.0743, 0.0699], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0194, 0.0245, 0.0212, 0.0205, 0.0247, 0.0249, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 09:43:04,713 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203103.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:43:05,261 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 09:43:07,153 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.348e+02 2.857e+02 3.355e+02 6.976e+02, threshold=5.714e+02, percent-clipped=1.0 +2023-02-07 09:43:17,239 INFO [train.py:901] (1/4) Epoch 26, batch 1050, loss[loss=0.1821, simple_loss=0.2725, pruned_loss=0.04581, over 8570.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2852, pruned_loss=0.06012, over 1601066.81 frames. ], batch size: 39, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:43:18,711 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 09:43:38,367 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2808, 1.7945, 4.3391, 1.9183, 2.5018, 4.9555, 5.0075, 4.2198], + device='cuda:1'), covar=tensor([0.1236, 0.1882, 0.0273, 0.2121, 0.1305, 0.0180, 0.0404, 0.0576], + device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0323, 0.0288, 0.0317, 0.0316, 0.0273, 0.0432, 0.0304], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 09:43:53,221 INFO [train.py:901] (1/4) Epoch 26, batch 1100, loss[loss=0.1794, simple_loss=0.2609, pruned_loss=0.04897, over 7657.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2848, pruned_loss=0.05962, over 1598198.19 frames. ], batch size: 19, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:44:06,850 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=203192.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:44:09,472 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8769, 1.6066, 1.7567, 1.5139, 1.0203, 1.6336, 1.7028, 1.4965], + device='cuda:1'), covar=tensor([0.0544, 0.1201, 0.1638, 0.1382, 0.0599, 0.1408, 0.0689, 0.0669], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0190, 0.0161, 0.0100, 0.0163, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 09:44:12,284 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7828, 1.4626, 3.3376, 1.5428, 2.3132, 3.6337, 3.6586, 3.1336], + device='cuda:1'), covar=tensor([0.1275, 0.1873, 0.0316, 0.1968, 0.1087, 0.0209, 0.0570, 0.0510], + device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0323, 0.0287, 0.0315, 0.0315, 0.0272, 0.0430, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 09:44:16,728 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.511e+02 2.912e+02 3.711e+02 8.666e+02, threshold=5.824e+02, percent-clipped=4.0 +2023-02-07 09:44:27,590 INFO [train.py:901] (1/4) Epoch 26, batch 1150, loss[loss=0.2177, simple_loss=0.3131, pruned_loss=0.06113, over 8327.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2854, pruned_loss=0.06021, over 1602368.69 frames. ], batch size: 25, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:44:27,626 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 09:45:02,743 INFO [train.py:901] (1/4) Epoch 26, batch 1200, loss[loss=0.1669, simple_loss=0.2518, pruned_loss=0.04093, over 7813.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2846, pruned_loss=0.05953, over 1604084.58 frames. ], batch size: 20, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:45:27,265 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203307.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:45:27,739 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.402e+02 2.806e+02 3.306e+02 6.331e+02, threshold=5.612e+02, percent-clipped=2.0 +2023-02-07 09:45:37,091 INFO [train.py:901] (1/4) Epoch 26, batch 1250, loss[loss=0.1808, simple_loss=0.2684, pruned_loss=0.04662, over 6902.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2853, pruned_loss=0.05949, over 1609281.81 frames. ], batch size: 74, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:45:49,713 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-07 09:46:12,725 INFO [train.py:901] (1/4) Epoch 26, batch 1300, loss[loss=0.164, simple_loss=0.2507, pruned_loss=0.03863, over 7916.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2844, pruned_loss=0.05894, over 1608155.88 frames. ], batch size: 20, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:46:31,897 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=203400.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:46:37,069 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.437e+02 2.919e+02 3.429e+02 9.499e+02, threshold=5.838e+02, percent-clipped=5.0 +2023-02-07 09:46:46,434 INFO [train.py:901] (1/4) Epoch 26, batch 1350, loss[loss=0.2012, simple_loss=0.2897, pruned_loss=0.0563, over 8313.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2872, pruned_loss=0.06051, over 1614326.83 frames. ], batch size: 25, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:47:04,180 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=203447.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:47:22,495 INFO [train.py:901] (1/4) Epoch 26, batch 1400, loss[loss=0.169, simple_loss=0.2644, pruned_loss=0.03685, over 8509.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2865, pruned_loss=0.05994, over 1616340.04 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:47:27,699 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-07 09:47:31,494 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203485.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:47:47,756 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.419e+02 2.906e+02 3.589e+02 5.599e+02, threshold=5.812e+02, percent-clipped=0.0 +2023-02-07 09:47:52,604 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203515.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:47:54,389 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 09:47:57,073 INFO [train.py:901] (1/4) Epoch 26, batch 1450, loss[loss=0.172, simple_loss=0.2558, pruned_loss=0.0441, over 8142.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2861, pruned_loss=0.05975, over 1617910.00 frames. ], batch size: 22, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:48:24,032 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203562.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:48:24,759 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203563.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:48:31,363 INFO [train.py:901] (1/4) Epoch 26, batch 1500, loss[loss=0.2357, simple_loss=0.3242, pruned_loss=0.07357, over 8105.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2855, pruned_loss=0.05978, over 1616926.77 frames. ], batch size: 23, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:48:32,172 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7554, 1.5819, 1.9443, 1.6839, 1.9239, 1.8023, 1.6812, 1.1555], + device='cuda:1'), covar=tensor([0.4357, 0.4038, 0.1768, 0.2980, 0.2043, 0.2580, 0.1601, 0.4146], + device='cuda:1'), in_proj_covar=tensor([0.0962, 0.1018, 0.0830, 0.0987, 0.1024, 0.0926, 0.0771, 0.0848], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 09:48:42,966 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203588.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:48:47,857 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 +2023-02-07 09:48:56,857 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.262e+02 2.668e+02 3.517e+02 8.500e+02, threshold=5.335e+02, percent-clipped=2.0 +2023-02-07 09:49:06,820 INFO [train.py:901] (1/4) Epoch 26, batch 1550, loss[loss=0.2019, simple_loss=0.2906, pruned_loss=0.05655, over 8316.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2855, pruned_loss=0.06009, over 1618225.09 frames. ], batch size: 25, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:49:40,405 INFO [train.py:901] (1/4) Epoch 26, batch 1600, loss[loss=0.1938, simple_loss=0.27, pruned_loss=0.05879, over 7934.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2853, pruned_loss=0.05942, over 1619964.65 frames. ], batch size: 20, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:49:41,261 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7908, 2.5592, 3.7107, 1.6373, 2.7508, 1.9683, 2.1098, 2.4685], + device='cuda:1'), covar=tensor([0.2039, 0.2304, 0.1101, 0.4902, 0.2051, 0.3882, 0.2312, 0.2826], + device='cuda:1'), in_proj_covar=tensor([0.0536, 0.0627, 0.0556, 0.0662, 0.0657, 0.0606, 0.0555, 0.0643], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:50:05,088 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.410e+02 3.021e+02 3.901e+02 1.362e+03, threshold=6.042e+02, percent-clipped=8.0 +2023-02-07 09:50:15,009 INFO [train.py:901] (1/4) Epoch 26, batch 1650, loss[loss=0.1747, simple_loss=0.2552, pruned_loss=0.0471, over 7662.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2855, pruned_loss=0.05981, over 1617145.50 frames. ], batch size: 19, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:50:18,603 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6897, 1.9679, 2.0393, 1.3366, 2.0731, 1.5850, 0.5670, 1.8801], + device='cuda:1'), covar=tensor([0.0606, 0.0363, 0.0339, 0.0624, 0.0482, 0.0955, 0.0920, 0.0320], + device='cuda:1'), in_proj_covar=tensor([0.0464, 0.0403, 0.0357, 0.0455, 0.0390, 0.0543, 0.0397, 0.0430], + device='cuda:1'), out_proj_covar=tensor([1.2327e-04, 1.0497e-04, 9.3347e-05, 1.1920e-04, 1.0216e-04, 1.5188e-04, + 1.0625e-04, 1.1293e-04], device='cuda:1') +2023-02-07 09:50:44,781 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203765.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:50:48,702 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203771.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:50:49,156 INFO [train.py:901] (1/4) Epoch 26, batch 1700, loss[loss=0.1774, simple_loss=0.2639, pruned_loss=0.04547, over 7549.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2856, pruned_loss=0.05996, over 1618857.04 frames. ], batch size: 18, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:50:57,394 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203784.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:05,481 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203796.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:14,291 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.578e+02 2.478e+02 3.017e+02 3.791e+02 8.735e+02, threshold=6.035e+02, percent-clipped=4.0 +2023-02-07 09:51:20,896 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203818.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:23,319 INFO [train.py:901] (1/4) Epoch 26, batch 1750, loss[loss=0.1783, simple_loss=0.2687, pruned_loss=0.04393, over 8094.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2851, pruned_loss=0.05925, over 1615542.75 frames. ], batch size: 21, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:51:28,103 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=203829.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:38,332 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:54,070 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:58,821 INFO [train.py:901] (1/4) Epoch 26, batch 1800, loss[loss=0.1606, simple_loss=0.2425, pruned_loss=0.03939, over 7547.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2834, pruned_loss=0.0585, over 1612136.99 frames. ], batch size: 18, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:52:06,722 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.53 vs. limit=5.0 +2023-02-07 09:52:15,983 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0595, 1.8575, 2.3578, 1.9832, 2.3008, 2.0509, 1.9458, 1.5726], + device='cuda:1'), covar=tensor([0.4294, 0.4141, 0.1714, 0.3086, 0.1984, 0.2703, 0.1652, 0.4044], + device='cuda:1'), in_proj_covar=tensor([0.0957, 0.1013, 0.0826, 0.0984, 0.1021, 0.0922, 0.0767, 0.0844], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 09:52:23,002 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.441e+02 2.799e+02 3.336e+02 4.977e+02, threshold=5.598e+02, percent-clipped=0.0 +2023-02-07 09:52:29,709 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203918.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:52:32,264 INFO [train.py:901] (1/4) Epoch 26, batch 1850, loss[loss=0.2237, simple_loss=0.3085, pruned_loss=0.0695, over 8469.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2827, pruned_loss=0.05823, over 1608887.21 frames. ], batch size: 25, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:52:47,699 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203944.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:53:07,707 INFO [train.py:901] (1/4) Epoch 26, batch 1900, loss[loss=0.1771, simple_loss=0.2644, pruned_loss=0.0449, over 7817.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2807, pruned_loss=0.0575, over 1603692.29 frames. ], batch size: 20, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:53:33,466 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.507e+02 3.073e+02 4.108e+02 9.647e+02, threshold=6.146e+02, percent-clipped=9.0 +2023-02-07 09:53:36,907 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 09:53:42,921 INFO [train.py:901] (1/4) Epoch 26, batch 1950, loss[loss=0.1928, simple_loss=0.2697, pruned_loss=0.05798, over 7655.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2804, pruned_loss=0.05713, over 1607137.23 frames. ], batch size: 19, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:53:49,445 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 09:54:07,151 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 09:54:11,412 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5152, 2.4950, 3.2693, 2.6537, 3.1292, 2.6051, 2.4433, 1.9484], + device='cuda:1'), covar=tensor([0.5459, 0.4980, 0.1982, 0.3657, 0.2511, 0.2896, 0.1789, 0.5444], + device='cuda:1'), in_proj_covar=tensor([0.0955, 0.1011, 0.0824, 0.0981, 0.1020, 0.0919, 0.0766, 0.0845], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 09:54:17,200 INFO [train.py:901] (1/4) Epoch 26, batch 2000, loss[loss=0.1824, simple_loss=0.2765, pruned_loss=0.04415, over 8807.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2808, pruned_loss=0.05708, over 1611175.72 frames. ], batch size: 40, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:54:34,357 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204095.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:54:43,742 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 2.388e+02 3.050e+02 3.690e+02 7.171e+02, threshold=6.101e+02, percent-clipped=4.0 +2023-02-07 09:54:44,463 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204109.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:54:53,076 INFO [train.py:901] (1/4) Epoch 26, batch 2050, loss[loss=0.1988, simple_loss=0.2749, pruned_loss=0.06141, over 7978.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2816, pruned_loss=0.05777, over 1606061.44 frames. ], batch size: 21, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:54:57,138 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204128.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:55:18,567 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8419, 3.7915, 3.5051, 1.7902, 3.4455, 3.5190, 3.3668, 3.3339], + device='cuda:1'), covar=tensor([0.0883, 0.0721, 0.1195, 0.4841, 0.0917, 0.1056, 0.1521, 0.0820], + device='cuda:1'), in_proj_covar=tensor([0.0536, 0.0453, 0.0437, 0.0551, 0.0436, 0.0457, 0.0432, 0.0397], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:55:26,603 INFO [train.py:901] (1/4) Epoch 26, batch 2100, loss[loss=0.1903, simple_loss=0.2802, pruned_loss=0.05017, over 8290.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2817, pruned_loss=0.05764, over 1606564.00 frames. ], batch size: 23, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:55:33,647 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204181.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:55:47,790 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204200.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:55:52,930 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.312e+02 2.797e+02 3.552e+02 6.063e+02, threshold=5.595e+02, percent-clipped=0.0 +2023-02-07 09:55:53,722 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204209.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:56:02,442 INFO [train.py:901] (1/4) Epoch 26, batch 2150, loss[loss=0.2098, simple_loss=0.2892, pruned_loss=0.06518, over 8340.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2813, pruned_loss=0.05723, over 1606097.90 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:56:03,976 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204224.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:56:04,625 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204225.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:56:17,184 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204243.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:56:29,891 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204262.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:56:36,294 INFO [train.py:901] (1/4) Epoch 26, batch 2200, loss[loss=0.1926, simple_loss=0.2734, pruned_loss=0.05588, over 8321.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2814, pruned_loss=0.05746, over 1609114.41 frames. ], batch size: 25, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:56:40,594 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.0487, 3.9491, 3.6927, 2.2513, 3.5337, 3.6848, 3.5791, 3.5157], + device='cuda:1'), covar=tensor([0.0816, 0.0681, 0.1003, 0.4223, 0.0916, 0.1107, 0.1364, 0.0882], + device='cuda:1'), in_proj_covar=tensor([0.0534, 0.0451, 0.0434, 0.0550, 0.0435, 0.0455, 0.0432, 0.0395], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 09:57:01,390 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.362e+02 3.099e+02 4.074e+02 1.599e+03, threshold=6.197e+02, percent-clipped=8.0 +2023-02-07 09:57:11,827 INFO [train.py:901] (1/4) Epoch 26, batch 2250, loss[loss=0.1941, simple_loss=0.2772, pruned_loss=0.05553, over 7412.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2823, pruned_loss=0.05789, over 1608445.65 frames. ], batch size: 17, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:57:13,374 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204324.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:57:29,318 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-07 09:57:46,491 INFO [train.py:901] (1/4) Epoch 26, batch 2300, loss[loss=0.1729, simple_loss=0.2625, pruned_loss=0.04166, over 7976.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2817, pruned_loss=0.05763, over 1607489.92 frames. ], batch size: 21, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:57:50,090 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204377.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:58:10,722 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.304e+02 2.813e+02 3.713e+02 7.684e+02, threshold=5.626e+02, percent-clipped=3.0 +2023-02-07 09:58:21,054 INFO [train.py:901] (1/4) Epoch 26, batch 2350, loss[loss=0.1836, simple_loss=0.2713, pruned_loss=0.04797, over 8246.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2825, pruned_loss=0.05824, over 1609482.61 frames. ], batch size: 22, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:58:33,859 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204439.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:58:57,219 INFO [train.py:901] (1/4) Epoch 26, batch 2400, loss[loss=0.1996, simple_loss=0.2868, pruned_loss=0.05623, over 8688.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2832, pruned_loss=0.05849, over 1612729.75 frames. ], batch size: 39, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:59:02,902 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204480.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:16,120 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204499.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:20,420 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204505.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:22,323 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.551e+02 2.904e+02 3.805e+02 7.023e+02, threshold=5.807e+02, percent-clipped=3.0 +2023-02-07 09:59:32,151 INFO [train.py:901] (1/4) Epoch 26, batch 2450, loss[loss=0.1958, simple_loss=0.2693, pruned_loss=0.06114, over 7929.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2827, pruned_loss=0.05819, over 1615035.52 frames. ], batch size: 20, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:59:33,770 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204524.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:34,351 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204525.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:56,610 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204554.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:00:08,715 INFO [train.py:901] (1/4) Epoch 26, batch 2500, loss[loss=0.1994, simple_loss=0.2873, pruned_loss=0.0558, over 8560.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2832, pruned_loss=0.05834, over 1615976.36 frames. ], batch size: 31, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:00:15,133 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204580.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:00:22,357 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7598, 5.8680, 5.1876, 2.3784, 5.2440, 5.5878, 5.3109, 5.3797], + device='cuda:1'), covar=tensor([0.0482, 0.0339, 0.0974, 0.4447, 0.0729, 0.0657, 0.1013, 0.0548], + device='cuda:1'), in_proj_covar=tensor([0.0532, 0.0449, 0.0434, 0.0547, 0.0433, 0.0454, 0.0432, 0.0394], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:00:31,998 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204605.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:00:33,808 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.483e+02 3.074e+02 3.585e+02 8.993e+02, threshold=6.148e+02, percent-clipped=7.0 +2023-02-07 10:00:35,164 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-07 10:00:43,171 INFO [train.py:901] (1/4) Epoch 26, batch 2550, loss[loss=0.2047, simple_loss=0.284, pruned_loss=0.06273, over 7978.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2842, pruned_loss=0.05901, over 1611074.89 frames. ], batch size: 21, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:00:50,562 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204633.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 10:00:54,541 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4272, 2.7725, 3.0584, 1.9859, 3.2821, 2.1279, 1.7363, 2.4740], + device='cuda:1'), covar=tensor([0.0940, 0.0439, 0.0394, 0.0767, 0.0488, 0.0889, 0.0935, 0.0558], + device='cuda:1'), in_proj_covar=tensor([0.0465, 0.0403, 0.0359, 0.0453, 0.0389, 0.0544, 0.0397, 0.0431], + device='cuda:1'), out_proj_covar=tensor([1.2365e-04, 1.0503e-04, 9.3730e-05, 1.1852e-04, 1.0189e-04, 1.5219e-04, + 1.0617e-04, 1.1351e-04], device='cuda:1') +2023-02-07 10:00:55,139 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204640.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:01:07,871 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204658.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 10:01:11,217 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5806, 1.9636, 2.9622, 1.4284, 2.2624, 1.8673, 1.6724, 2.2750], + device='cuda:1'), covar=tensor([0.1921, 0.2528, 0.0897, 0.4682, 0.1823, 0.3317, 0.2434, 0.2200], + device='cuda:1'), in_proj_covar=tensor([0.0535, 0.0623, 0.0556, 0.0659, 0.0655, 0.0605, 0.0552, 0.0640], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:01:18,440 INFO [train.py:901] (1/4) Epoch 26, batch 2600, loss[loss=0.2193, simple_loss=0.3036, pruned_loss=0.06755, over 8363.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2845, pruned_loss=0.05921, over 1612637.10 frames. ], batch size: 24, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:01:29,263 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1781, 1.5120, 1.7036, 1.4497, 1.0201, 1.5068, 1.7581, 1.6364], + device='cuda:1'), covar=tensor([0.0469, 0.1286, 0.1642, 0.1442, 0.0586, 0.1498, 0.0711, 0.0661], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0191, 0.0161, 0.0101, 0.0163, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 10:01:43,344 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.465e+02 3.094e+02 3.874e+02 9.576e+02, threshold=6.187e+02, percent-clipped=4.0 +2023-02-07 10:01:52,896 INFO [train.py:901] (1/4) Epoch 26, batch 2650, loss[loss=0.1862, simple_loss=0.2713, pruned_loss=0.05057, over 7802.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.285, pruned_loss=0.05898, over 1616895.07 frames. ], batch size: 20, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:01:57,927 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2100, 2.0866, 2.6236, 2.1553, 2.5750, 2.2868, 2.0969, 1.3887], + device='cuda:1'), covar=tensor([0.6008, 0.5086, 0.2050, 0.4284, 0.2812, 0.3383, 0.2202, 0.5835], + device='cuda:1'), in_proj_covar=tensor([0.0958, 0.1012, 0.0824, 0.0981, 0.1018, 0.0920, 0.0767, 0.0844], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:02:19,919 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0322, 1.9196, 2.4342, 2.0306, 2.3306, 2.1292, 1.9759, 1.2233], + device='cuda:1'), covar=tensor([0.6415, 0.5312, 0.1937, 0.3876, 0.2760, 0.3449, 0.2300, 0.5316], + device='cuda:1'), in_proj_covar=tensor([0.0955, 0.1009, 0.0821, 0.0977, 0.1015, 0.0918, 0.0764, 0.0841], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:02:28,025 INFO [train.py:901] (1/4) Epoch 26, batch 2700, loss[loss=0.219, simple_loss=0.3088, pruned_loss=0.06458, over 8346.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2847, pruned_loss=0.05904, over 1616152.82 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:02:36,921 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0670, 1.4543, 1.7521, 1.3370, 0.9703, 1.5160, 1.7397, 1.8571], + device='cuda:1'), covar=tensor([0.0493, 0.1258, 0.1674, 0.1447, 0.0589, 0.1439, 0.0679, 0.0588], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0190, 0.0161, 0.0101, 0.0163, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 10:02:53,797 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.359e+02 2.865e+02 3.674e+02 6.992e+02, threshold=5.730e+02, percent-clipped=1.0 +2023-02-07 10:02:55,417 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204810.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:03:04,000 INFO [train.py:901] (1/4) Epoch 26, batch 2750, loss[loss=0.2233, simple_loss=0.303, pruned_loss=0.07178, over 7673.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2839, pruned_loss=0.05866, over 1616958.47 frames. ], batch size: 19, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:03:13,266 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204835.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:03:16,006 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204839.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:03:38,129 INFO [train.py:901] (1/4) Epoch 26, batch 2800, loss[loss=0.2357, simple_loss=0.312, pruned_loss=0.07964, over 8334.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2844, pruned_loss=0.05871, over 1616875.72 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:03:55,281 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204896.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:04:04,626 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.285e+02 3.108e+02 3.828e+02 9.944e+02, threshold=6.216e+02, percent-clipped=6.0 +2023-02-07 10:04:13,928 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204921.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:04:14,420 INFO [train.py:901] (1/4) Epoch 26, batch 2850, loss[loss=0.1916, simple_loss=0.2818, pruned_loss=0.05067, over 8484.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2839, pruned_loss=0.0586, over 1616738.28 frames. ], batch size: 25, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:04:23,876 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9751, 1.4525, 3.4590, 1.7197, 2.4756, 3.7564, 3.8521, 3.2008], + device='cuda:1'), covar=tensor([0.1238, 0.1962, 0.0280, 0.1925, 0.1008, 0.0223, 0.0538, 0.0542], + device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0327, 0.0290, 0.0319, 0.0321, 0.0276, 0.0435, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 10:04:48,513 INFO [train.py:901] (1/4) Epoch 26, batch 2900, loss[loss=0.2, simple_loss=0.2913, pruned_loss=0.05441, over 8135.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2845, pruned_loss=0.05885, over 1615200.73 frames. ], batch size: 22, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:04:59,572 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.30 vs. limit=5.0 +2023-02-07 10:05:00,980 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 10:05:13,373 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.393e+02 3.052e+02 3.991e+02 9.487e+02, threshold=6.105e+02, percent-clipped=5.0 +2023-02-07 10:05:20,484 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 10:05:23,876 INFO [train.py:901] (1/4) Epoch 26, batch 2950, loss[loss=0.222, simple_loss=0.3107, pruned_loss=0.06661, over 8489.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2844, pruned_loss=0.05885, over 1616461.47 frames. ], batch size: 29, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:05:58,484 INFO [train.py:901] (1/4) Epoch 26, batch 3000, loss[loss=0.1713, simple_loss=0.2516, pruned_loss=0.04547, over 7437.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.284, pruned_loss=0.05853, over 1613808.37 frames. ], batch size: 17, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:05:58,484 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 10:06:08,067 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.7555, 1.4852, 3.9059, 1.6528, 3.4709, 3.2676, 3.6006, 3.5164], + device='cuda:1'), covar=tensor([0.0760, 0.4715, 0.0564, 0.4283, 0.1156, 0.1095, 0.0702, 0.0714], + device='cuda:1'), in_proj_covar=tensor([0.0665, 0.0662, 0.0728, 0.0650, 0.0742, 0.0629, 0.0629, 0.0706], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:06:11,420 INFO [train.py:935] (1/4) Epoch 26, validation: loss=0.1716, simple_loss=0.2713, pruned_loss=0.03593, over 944034.00 frames. +2023-02-07 10:06:11,421 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 10:06:31,073 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-02-07 10:06:36,703 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 2.246e+02 2.785e+02 3.735e+02 7.523e+02, threshold=5.571e+02, percent-clipped=3.0 +2023-02-07 10:06:46,005 INFO [train.py:901] (1/4) Epoch 26, batch 3050, loss[loss=0.1723, simple_loss=0.262, pruned_loss=0.04124, over 8693.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2839, pruned_loss=0.05817, over 1616558.76 frames. ], batch size: 39, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:06:49,437 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205127.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:07:22,793 INFO [train.py:901] (1/4) Epoch 26, batch 3100, loss[loss=0.2313, simple_loss=0.3053, pruned_loss=0.07861, over 8447.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2837, pruned_loss=0.05792, over 1615988.13 frames. ], batch size: 29, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:07:30,228 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=205183.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:07:48,159 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.327e+02 2.997e+02 4.038e+02 1.256e+03, threshold=5.993e+02, percent-clipped=7.0 +2023-02-07 10:07:53,680 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205216.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:07:57,552 INFO [train.py:901] (1/4) Epoch 26, batch 3150, loss[loss=0.1818, simple_loss=0.2606, pruned_loss=0.0515, over 7451.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2843, pruned_loss=0.05845, over 1617133.55 frames. ], batch size: 17, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:08:33,479 INFO [train.py:901] (1/4) Epoch 26, batch 3200, loss[loss=0.2372, simple_loss=0.3161, pruned_loss=0.0792, over 8473.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2834, pruned_loss=0.05794, over 1614526.26 frames. ], batch size: 25, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:08:52,245 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205298.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:08:57,569 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.9729, 6.1705, 5.3802, 2.8264, 5.4871, 5.7065, 5.6220, 5.5612], + device='cuda:1'), covar=tensor([0.0506, 0.0298, 0.0863, 0.3954, 0.0707, 0.0669, 0.1089, 0.0505], + device='cuda:1'), in_proj_covar=tensor([0.0534, 0.0449, 0.0437, 0.0549, 0.0432, 0.0455, 0.0431, 0.0394], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:08:58,813 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.527e+02 3.010e+02 3.735e+02 6.895e+02, threshold=6.021e+02, percent-clipped=2.0 +2023-02-07 10:09:09,105 INFO [train.py:901] (1/4) Epoch 26, batch 3250, loss[loss=0.2036, simple_loss=0.2902, pruned_loss=0.05852, over 8598.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2827, pruned_loss=0.05782, over 1610640.45 frames. ], batch size: 39, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:09:43,189 INFO [train.py:901] (1/4) Epoch 26, batch 3300, loss[loss=0.206, simple_loss=0.2931, pruned_loss=0.05948, over 8509.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2831, pruned_loss=0.05793, over 1610409.57 frames. ], batch size: 26, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:09:52,943 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9533, 1.9678, 1.8160, 2.5582, 1.0543, 1.5511, 1.9136, 2.1064], + device='cuda:1'), covar=tensor([0.0758, 0.0788, 0.0887, 0.0439, 0.1153, 0.1360, 0.0795, 0.0717], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0192, 0.0243, 0.0210, 0.0203, 0.0245, 0.0248, 0.0204], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 10:10:10,316 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.313e+02 2.653e+02 3.358e+02 9.214e+02, threshold=5.305e+02, percent-clipped=4.0 +2023-02-07 10:10:16,828 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4392, 1.8292, 2.6332, 1.3425, 2.0037, 1.8836, 1.4638, 2.1101], + device='cuda:1'), covar=tensor([0.1906, 0.2571, 0.0868, 0.4603, 0.1917, 0.3146, 0.2584, 0.2054], + device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0626, 0.0559, 0.0663, 0.0658, 0.0607, 0.0557, 0.0640], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:10:20,068 INFO [train.py:901] (1/4) Epoch 26, batch 3350, loss[loss=0.1893, simple_loss=0.272, pruned_loss=0.05326, over 8293.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2825, pruned_loss=0.05768, over 1608964.69 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:10:54,102 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=205471.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:10:54,720 INFO [train.py:901] (1/4) Epoch 26, batch 3400, loss[loss=0.2411, simple_loss=0.3133, pruned_loss=0.08443, over 8514.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2827, pruned_loss=0.05751, over 1608307.18 frames. ], batch size: 31, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:11:02,813 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-07 10:11:03,621 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2142, 1.5538, 4.4080, 2.1204, 2.4605, 5.1038, 5.2175, 4.4212], + device='cuda:1'), covar=tensor([0.1256, 0.1949, 0.0228, 0.1857, 0.1246, 0.0185, 0.0342, 0.0597], + device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0327, 0.0290, 0.0319, 0.0321, 0.0276, 0.0435, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 10:11:20,313 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.409e+02 2.883e+02 3.635e+02 7.106e+02, threshold=5.767e+02, percent-clipped=3.0 +2023-02-07 10:11:30,468 INFO [train.py:901] (1/4) Epoch 26, batch 3450, loss[loss=0.2123, simple_loss=0.2988, pruned_loss=0.06291, over 8486.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2827, pruned_loss=0.05761, over 1607958.23 frames. ], batch size: 28, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:11:53,164 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205554.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:11:57,167 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=205560.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:12:05,234 INFO [train.py:901] (1/4) Epoch 26, batch 3500, loss[loss=0.177, simple_loss=0.2563, pruned_loss=0.04889, over 8031.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2826, pruned_loss=0.05759, over 1606364.14 frames. ], batch size: 20, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:12:10,360 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205579.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:12:15,154 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:12:21,534 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6101, 2.0768, 3.0779, 1.4567, 2.2593, 1.9938, 1.6686, 2.4055], + device='cuda:1'), covar=tensor([0.1912, 0.2632, 0.0975, 0.4788, 0.2118, 0.3352, 0.2467, 0.2457], + device='cuda:1'), in_proj_covar=tensor([0.0536, 0.0626, 0.0558, 0.0662, 0.0658, 0.0605, 0.0557, 0.0640], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:12:24,733 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 10:12:30,113 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.208e+02 2.714e+02 3.358e+02 5.744e+02, threshold=5.428e+02, percent-clipped=0.0 +2023-02-07 10:12:35,786 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 10:12:39,510 INFO [train.py:901] (1/4) Epoch 26, batch 3550, loss[loss=0.1658, simple_loss=0.2577, pruned_loss=0.03698, over 7970.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2823, pruned_loss=0.05727, over 1610058.30 frames. ], batch size: 21, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:13:15,373 INFO [train.py:901] (1/4) Epoch 26, batch 3600, loss[loss=0.1725, simple_loss=0.2509, pruned_loss=0.04699, over 7247.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2823, pruned_loss=0.0576, over 1611458.30 frames. ], batch size: 16, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:13:16,199 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205673.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:13:17,473 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205675.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:13:39,661 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.295e+02 2.882e+02 3.730e+02 8.207e+02, threshold=5.763e+02, percent-clipped=6.0 +2023-02-07 10:13:49,105 INFO [train.py:901] (1/4) Epoch 26, batch 3650, loss[loss=0.1977, simple_loss=0.2721, pruned_loss=0.06164, over 7966.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2824, pruned_loss=0.05744, over 1610662.65 frames. ], batch size: 21, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:14:18,433 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205762.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:14:24,801 INFO [train.py:901] (1/4) Epoch 26, batch 3700, loss[loss=0.1816, simple_loss=0.2738, pruned_loss=0.04471, over 8504.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2832, pruned_loss=0.05807, over 1616570.52 frames. ], batch size: 29, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:14:27,615 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 10:14:49,669 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.377e+02 2.968e+02 3.727e+02 1.221e+03, threshold=5.937e+02, percent-clipped=5.0 +2023-02-07 10:14:59,197 INFO [train.py:901] (1/4) Epoch 26, batch 3750, loss[loss=0.17, simple_loss=0.2677, pruned_loss=0.03614, over 8308.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.283, pruned_loss=0.05824, over 1614364.45 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:15:12,920 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205842.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:15:31,310 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205867.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:15:34,019 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5132, 1.3964, 1.8474, 1.3109, 1.1273, 1.8071, 0.1904, 1.1789], + device='cuda:1'), covar=tensor([0.1291, 0.1166, 0.0349, 0.0809, 0.2458, 0.0422, 0.1816, 0.1186], + device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0202, 0.0132, 0.0221, 0.0274, 0.0144, 0.0171, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 10:15:34,439 INFO [train.py:901] (1/4) Epoch 26, batch 3800, loss[loss=0.2189, simple_loss=0.3013, pruned_loss=0.06819, over 8369.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2817, pruned_loss=0.05794, over 1611253.51 frames. ], batch size: 24, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:15:59,216 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 2.381e+02 2.847e+02 3.364e+02 6.986e+02, threshold=5.694e+02, percent-clipped=1.0 +2023-02-07 10:15:59,375 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205908.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:16:08,803 INFO [train.py:901] (1/4) Epoch 26, batch 3850, loss[loss=0.1954, simple_loss=0.275, pruned_loss=0.05788, over 8290.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2817, pruned_loss=0.05775, over 1614691.25 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:16:15,133 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205931.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:16:20,582 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3748, 1.6052, 4.3330, 2.0534, 2.6217, 5.0326, 5.1242, 4.3113], + device='cuda:1'), covar=tensor([0.1240, 0.2007, 0.0290, 0.2029, 0.1211, 0.0177, 0.0475, 0.0552], + device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0328, 0.0291, 0.0321, 0.0322, 0.0279, 0.0438, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 10:16:23,272 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9464, 2.6706, 4.0816, 1.7106, 3.1932, 2.3773, 2.0718, 3.0596], + device='cuda:1'), covar=tensor([0.1816, 0.2334, 0.0840, 0.4410, 0.1693, 0.3089, 0.2262, 0.2240], + device='cuda:1'), in_proj_covar=tensor([0.0537, 0.0626, 0.0558, 0.0662, 0.0657, 0.0605, 0.0557, 0.0641], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:16:29,158 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 10:16:32,101 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205956.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:16:42,785 INFO [train.py:901] (1/4) Epoch 26, batch 3900, loss[loss=0.1948, simple_loss=0.2803, pruned_loss=0.05461, over 8135.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2821, pruned_loss=0.05853, over 1610794.59 frames. ], batch size: 22, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:17:09,986 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.608e+02 2.441e+02 2.892e+02 3.706e+02 7.796e+02, threshold=5.785e+02, percent-clipped=3.0 +2023-02-07 10:17:16,700 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=206017.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:17:18,466 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-07 10:17:20,038 INFO [train.py:901] (1/4) Epoch 26, batch 3950, loss[loss=0.174, simple_loss=0.2507, pruned_loss=0.04868, over 7260.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2822, pruned_loss=0.05844, over 1609518.26 frames. ], batch size: 16, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:17:53,922 INFO [train.py:901] (1/4) Epoch 26, batch 4000, loss[loss=0.1606, simple_loss=0.2538, pruned_loss=0.03373, over 8149.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2833, pruned_loss=0.05852, over 1614342.51 frames. ], batch size: 22, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:17:55,452 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206074.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:18:18,672 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=206106.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:18:19,947 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.407e+02 2.986e+02 3.556e+02 8.558e+02, threshold=5.971e+02, percent-clipped=6.0 +2023-02-07 10:18:29,512 INFO [train.py:901] (1/4) Epoch 26, batch 4050, loss[loss=0.2361, simple_loss=0.3264, pruned_loss=0.07287, over 8524.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2833, pruned_loss=0.05843, over 1609854.14 frames. ], batch size: 48, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:18:37,175 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206132.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:19:03,800 INFO [train.py:901] (1/4) Epoch 26, batch 4100, loss[loss=0.1644, simple_loss=0.2443, pruned_loss=0.04225, over 7928.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2822, pruned_loss=0.05722, over 1612157.06 frames. ], batch size: 20, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:19:28,870 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.376e+02 2.755e+02 3.418e+02 9.873e+02, threshold=5.510e+02, percent-clipped=4.0 +2023-02-07 10:19:31,067 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6811, 1.8908, 2.0515, 1.8392, 1.2397, 1.8915, 2.2657, 2.1658], + device='cuda:1'), covar=tensor([0.0513, 0.1142, 0.1616, 0.1348, 0.0627, 0.1333, 0.0670, 0.0556], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0153, 0.0189, 0.0160, 0.0100, 0.0162, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 10:19:34,493 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206215.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:19:39,251 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206221.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:19:39,711 INFO [train.py:901] (1/4) Epoch 26, batch 4150, loss[loss=0.189, simple_loss=0.2678, pruned_loss=0.05511, over 7812.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2819, pruned_loss=0.05722, over 1615473.61 frames. ], batch size: 19, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:19:52,402 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.47 vs. limit=5.0 +2023-02-07 10:20:00,463 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 10:20:00,779 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=206252.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:20:10,905 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8359, 3.8069, 3.4547, 1.9813, 3.4174, 3.4625, 3.3690, 3.3524], + device='cuda:1'), covar=tensor([0.0902, 0.0666, 0.1230, 0.4470, 0.1026, 0.1184, 0.1512, 0.0931], + device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0455, 0.0441, 0.0553, 0.0436, 0.0459, 0.0436, 0.0400], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:20:14,202 INFO [train.py:901] (1/4) Epoch 26, batch 4200, loss[loss=0.1854, simple_loss=0.2692, pruned_loss=0.05076, over 7906.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2803, pruned_loss=0.05656, over 1611365.40 frames. ], batch size: 20, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:20:22,977 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 10:20:38,388 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 2.335e+02 2.968e+02 3.755e+02 9.805e+02, threshold=5.936e+02, percent-clipped=3.0 +2023-02-07 10:20:44,922 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 10:20:49,138 INFO [train.py:901] (1/4) Epoch 26, batch 4250, loss[loss=0.2188, simple_loss=0.2991, pruned_loss=0.06924, over 8462.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2806, pruned_loss=0.05661, over 1612775.77 frames. ], batch size: 27, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:20:54,076 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.8283, 3.7981, 3.4560, 1.7794, 3.3513, 3.5020, 3.3954, 3.3651], + device='cuda:1'), covar=tensor([0.0933, 0.0688, 0.1203, 0.5048, 0.1095, 0.1297, 0.1492, 0.0936], + device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0457, 0.0443, 0.0556, 0.0439, 0.0461, 0.0438, 0.0401], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:21:21,338 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206367.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:21:24,640 INFO [train.py:901] (1/4) Epoch 26, batch 4300, loss[loss=0.1888, simple_loss=0.2752, pruned_loss=0.05118, over 7972.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2811, pruned_loss=0.05679, over 1613072.00 frames. ], batch size: 21, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:21:31,160 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 10:21:35,882 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206388.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:21:50,294 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.331e+02 2.890e+02 3.800e+02 6.492e+02, threshold=5.781e+02, percent-clipped=2.0 +2023-02-07 10:21:53,273 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206413.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:21:56,606 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=206418.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:21:59,348 INFO [train.py:901] (1/4) Epoch 26, batch 4350, loss[loss=0.1753, simple_loss=0.2706, pruned_loss=0.04005, over 8596.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2814, pruned_loss=0.05689, over 1615128.22 frames. ], batch size: 31, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:22:18,982 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 10:22:34,790 INFO [train.py:901] (1/4) Epoch 26, batch 4400, loss[loss=0.2207, simple_loss=0.3031, pruned_loss=0.06909, over 8113.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2835, pruned_loss=0.05767, over 1619537.96 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:22:38,347 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206477.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:22:55,677 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206502.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:23:00,163 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.619e+02 3.000e+02 3.925e+02 8.429e+02, threshold=6.000e+02, percent-clipped=7.0 +2023-02-07 10:23:00,196 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 10:23:08,813 INFO [train.py:901] (1/4) Epoch 26, batch 4450, loss[loss=0.1848, simple_loss=0.274, pruned_loss=0.04776, over 8088.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.283, pruned_loss=0.05763, over 1619132.73 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:23:16,247 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206533.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:23:16,908 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5853, 1.5511, 4.7633, 1.7815, 4.2388, 4.0460, 4.3282, 4.1948], + device='cuda:1'), covar=tensor([0.0566, 0.4868, 0.0504, 0.4098, 0.1126, 0.0879, 0.0551, 0.0652], + device='cuda:1'), in_proj_covar=tensor([0.0661, 0.0655, 0.0724, 0.0648, 0.0735, 0.0622, 0.0623, 0.0696], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:23:20,342 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-07 10:23:34,157 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=206559.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:23:44,263 INFO [train.py:901] (1/4) Epoch 26, batch 4500, loss[loss=0.2043, simple_loss=0.2934, pruned_loss=0.05759, over 8246.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2829, pruned_loss=0.058, over 1617267.83 frames. ], batch size: 24, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:23:55,206 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 10:23:59,317 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1140, 2.2005, 1.9743, 2.7155, 1.2599, 1.7282, 2.0107, 2.1935], + device='cuda:1'), covar=tensor([0.0743, 0.0802, 0.0803, 0.0383, 0.1070, 0.1248, 0.0804, 0.0899], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0194, 0.0244, 0.0212, 0.0203, 0.0246, 0.0249, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 10:24:10,073 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 2.363e+02 2.961e+02 3.499e+02 6.135e+02, threshold=5.921e+02, percent-clipped=1.0 +2023-02-07 10:24:18,689 INFO [train.py:901] (1/4) Epoch 26, batch 4550, loss[loss=0.1882, simple_loss=0.2715, pruned_loss=0.05243, over 7658.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2824, pruned_loss=0.05804, over 1614022.87 frames. ], batch size: 19, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:24:19,486 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206623.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:24:35,788 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206648.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:24:51,534 INFO [train.py:901] (1/4) Epoch 26, batch 4600, loss[loss=0.2123, simple_loss=0.2911, pruned_loss=0.06675, over 8597.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2825, pruned_loss=0.05855, over 1610691.11 frames. ], batch size: 31, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:24:53,057 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206674.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:25:18,508 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.342e+02 2.811e+02 3.625e+02 9.770e+02, threshold=5.622e+02, percent-clipped=5.0 +2023-02-07 10:25:20,162 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6313, 2.6998, 1.8867, 2.4200, 2.1762, 1.6543, 2.1202, 2.2681], + device='cuda:1'), covar=tensor([0.1586, 0.0405, 0.1232, 0.0682, 0.0811, 0.1590, 0.1190, 0.1067], + device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0241, 0.0342, 0.0314, 0.0304, 0.0348, 0.0350, 0.0325], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 10:25:23,279 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-07 10:25:28,328 INFO [train.py:901] (1/4) Epoch 26, batch 4650, loss[loss=0.201, simple_loss=0.2878, pruned_loss=0.05709, over 7668.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2826, pruned_loss=0.05862, over 1609756.82 frames. ], batch size: 19, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:25:40,977 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 10:26:02,070 INFO [train.py:901] (1/4) Epoch 26, batch 4700, loss[loss=0.2214, simple_loss=0.3126, pruned_loss=0.06511, over 8318.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2823, pruned_loss=0.0585, over 1610291.84 frames. ], batch size: 25, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:26:08,424 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0451, 1.2658, 1.1991, 0.7332, 1.1944, 1.0976, 0.0625, 1.2042], + device='cuda:1'), covar=tensor([0.0514, 0.0441, 0.0417, 0.0703, 0.0512, 0.1070, 0.0997, 0.0428], + device='cuda:1'), in_proj_covar=tensor([0.0466, 0.0402, 0.0360, 0.0455, 0.0389, 0.0547, 0.0399, 0.0434], + device='cuda:1'), out_proj_covar=tensor([1.2370e-04, 1.0477e-04, 9.4011e-05, 1.1921e-04, 1.0177e-04, 1.5321e-04, + 1.0686e-04, 1.1408e-04], device='cuda:1') +2023-02-07 10:26:13,620 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206789.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:26:28,922 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.506e+02 2.890e+02 3.298e+02 6.611e+02, threshold=5.779e+02, percent-clipped=3.0 +2023-02-07 10:26:32,508 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206814.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:26:34,595 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2501, 2.0833, 2.7680, 2.2919, 2.7177, 2.2985, 2.1192, 1.5875], + device='cuda:1'), covar=tensor([0.5675, 0.5294, 0.2062, 0.4122, 0.2744, 0.3320, 0.2068, 0.5900], + device='cuda:1'), in_proj_covar=tensor([0.0952, 0.1003, 0.0824, 0.0979, 0.1014, 0.0917, 0.0764, 0.0838], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:26:37,690 INFO [train.py:901] (1/4) Epoch 26, batch 4750, loss[loss=0.164, simple_loss=0.25, pruned_loss=0.03898, over 8090.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2833, pruned_loss=0.0588, over 1608961.79 frames. ], batch size: 21, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:26:53,378 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 10:26:55,377 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 10:26:58,849 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206852.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:27:00,991 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-07 10:27:12,097 INFO [train.py:901] (1/4) Epoch 26, batch 4800, loss[loss=0.17, simple_loss=0.248, pruned_loss=0.04604, over 7547.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2821, pruned_loss=0.05874, over 1602440.76 frames. ], batch size: 18, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:27:21,153 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-07 10:27:37,291 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 2.410e+02 2.886e+02 3.541e+02 7.542e+02, threshold=5.772e+02, percent-clipped=6.0 +2023-02-07 10:27:46,701 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 10:27:47,384 INFO [train.py:901] (1/4) Epoch 26, batch 4850, loss[loss=0.2375, simple_loss=0.3109, pruned_loss=0.08212, over 8487.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2831, pruned_loss=0.05888, over 1609188.03 frames. ], batch size: 28, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:27:52,975 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206930.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:28:10,405 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206955.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:28:21,655 INFO [train.py:901] (1/4) Epoch 26, batch 4900, loss[loss=0.2034, simple_loss=0.2885, pruned_loss=0.05916, over 8604.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2839, pruned_loss=0.05929, over 1611926.18 frames. ], batch size: 49, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:28:25,971 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.05 vs. limit=5.0 +2023-02-07 10:28:46,039 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.535e+02 3.142e+02 3.836e+02 8.051e+02, threshold=6.285e+02, percent-clipped=2.0 +2023-02-07 10:28:55,279 INFO [train.py:901] (1/4) Epoch 26, batch 4950, loss[loss=0.2149, simple_loss=0.2886, pruned_loss=0.07057, over 6828.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2847, pruned_loss=0.05987, over 1608511.56 frames. ], batch size: 71, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:29:29,754 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8521, 1.7347, 2.4595, 1.4290, 1.3240, 2.3788, 0.4270, 1.4583], + device='cuda:1'), covar=tensor([0.1378, 0.1127, 0.0304, 0.1300, 0.2407, 0.0390, 0.1953, 0.1419], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0202, 0.0131, 0.0220, 0.0273, 0.0142, 0.0170, 0.0195], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 10:29:32,342 INFO [train.py:901] (1/4) Epoch 26, batch 5000, loss[loss=0.1943, simple_loss=0.2727, pruned_loss=0.05802, over 8283.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2847, pruned_loss=0.05972, over 1607458.16 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:29:42,773 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1205, 2.2835, 1.9898, 2.9335, 1.3150, 1.7045, 2.1123, 2.2400], + device='cuda:1'), covar=tensor([0.0694, 0.0732, 0.0769, 0.0324, 0.1119, 0.1250, 0.0818, 0.0772], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0194, 0.0245, 0.0213, 0.0204, 0.0247, 0.0250, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 10:29:57,379 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.413e+02 2.985e+02 3.933e+02 1.062e+03, threshold=5.970e+02, percent-clipped=3.0 +2023-02-07 10:30:06,454 INFO [train.py:901] (1/4) Epoch 26, batch 5050, loss[loss=0.1777, simple_loss=0.2806, pruned_loss=0.0374, over 8534.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2839, pruned_loss=0.05942, over 1607283.65 frames. ], batch size: 28, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:30:22,941 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3926, 2.0783, 2.6497, 2.2229, 2.6214, 2.3804, 2.1949, 1.5439], + device='cuda:1'), covar=tensor([0.5571, 0.5159, 0.1993, 0.4071, 0.2713, 0.3058, 0.1930, 0.5581], + device='cuda:1'), in_proj_covar=tensor([0.0957, 0.1005, 0.0825, 0.0982, 0.1016, 0.0919, 0.0764, 0.0843], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:30:24,762 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 10:30:42,628 INFO [train.py:901] (1/4) Epoch 26, batch 5100, loss[loss=0.1563, simple_loss=0.2408, pruned_loss=0.03585, over 7529.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2838, pruned_loss=0.05921, over 1606639.05 frames. ], batch size: 18, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:30:58,149 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207194.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:30:59,446 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=207196.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:31:08,240 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.087e+02 2.633e+02 3.622e+02 6.552e+02, threshold=5.265e+02, percent-clipped=1.0 +2023-02-07 10:31:16,944 INFO [train.py:901] (1/4) Epoch 26, batch 5150, loss[loss=0.1677, simple_loss=0.2451, pruned_loss=0.04519, over 7523.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2837, pruned_loss=0.05928, over 1601496.59 frames. ], batch size: 18, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:31:52,800 INFO [train.py:901] (1/4) Epoch 26, batch 5200, loss[loss=0.1836, simple_loss=0.2757, pruned_loss=0.04577, over 8487.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2836, pruned_loss=0.05911, over 1606185.17 frames. ], batch size: 49, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:32:17,988 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.585e+02 3.464e+02 4.468e+02 1.375e+03, threshold=6.928e+02, percent-clipped=16.0 +2023-02-07 10:32:18,903 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.17 vs. limit=5.0 +2023-02-07 10:32:19,451 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207311.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:32:19,962 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 10:32:26,696 INFO [train.py:901] (1/4) Epoch 26, batch 5250, loss[loss=0.187, simple_loss=0.2845, pruned_loss=0.04477, over 8338.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2834, pruned_loss=0.05878, over 1604264.26 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:32:48,497 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0639, 1.7885, 3.3726, 1.4498, 2.3913, 3.6848, 3.8618, 3.1693], + device='cuda:1'), covar=tensor([0.1222, 0.1772, 0.0335, 0.2316, 0.1108, 0.0255, 0.0563, 0.0563], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0321, 0.0286, 0.0314, 0.0314, 0.0273, 0.0429, 0.0301], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 10:33:00,346 INFO [train.py:901] (1/4) Epoch 26, batch 5300, loss[loss=0.2358, simple_loss=0.3185, pruned_loss=0.0765, over 8195.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2834, pruned_loss=0.05816, over 1610061.01 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:33:27,787 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.392e+02 2.913e+02 3.782e+02 6.658e+02, threshold=5.826e+02, percent-clipped=0.0 +2023-02-07 10:33:36,852 INFO [train.py:901] (1/4) Epoch 26, batch 5350, loss[loss=0.2082, simple_loss=0.298, pruned_loss=0.05918, over 8353.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2829, pruned_loss=0.0579, over 1611460.88 frames. ], batch size: 24, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:34:10,290 INFO [train.py:901] (1/4) Epoch 26, batch 5400, loss[loss=0.2093, simple_loss=0.3048, pruned_loss=0.05687, over 8296.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2843, pruned_loss=0.05834, over 1619414.50 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:34:37,323 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.344e+02 3.061e+02 4.157e+02 9.885e+02, threshold=6.122e+02, percent-clipped=8.0 +2023-02-07 10:34:46,158 INFO [train.py:901] (1/4) Epoch 26, batch 5450, loss[loss=0.2118, simple_loss=0.2998, pruned_loss=0.06189, over 8322.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2837, pruned_loss=0.05824, over 1615300.87 frames. ], batch size: 25, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:34:47,014 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8029, 1.5515, 1.7144, 1.4212, 0.9717, 1.5232, 1.6871, 1.4300], + device='cuda:1'), covar=tensor([0.0580, 0.1207, 0.1663, 0.1489, 0.0606, 0.1448, 0.0726, 0.0673], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0154, 0.0190, 0.0161, 0.0101, 0.0163, 0.0113, 0.0146], + device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 10:34:57,839 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=207538.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:35:00,781 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8697, 1.6767, 1.9994, 1.7964, 1.8966, 1.9451, 1.8054, 0.8182], + device='cuda:1'), covar=tensor([0.6068, 0.4824, 0.2176, 0.3790, 0.2771, 0.3399, 0.2041, 0.5598], + device='cuda:1'), in_proj_covar=tensor([0.0959, 0.1009, 0.0827, 0.0984, 0.1019, 0.0921, 0.0765, 0.0844], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:35:06,590 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 10:35:08,192 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6004, 2.9152, 3.2500, 1.8801, 3.4471, 2.3037, 1.6657, 2.4629], + device='cuda:1'), covar=tensor([0.0762, 0.0366, 0.0265, 0.0808, 0.0467, 0.0777, 0.0964, 0.0561], + device='cuda:1'), in_proj_covar=tensor([0.0465, 0.0404, 0.0359, 0.0456, 0.0390, 0.0548, 0.0401, 0.0434], + device='cuda:1'), out_proj_covar=tensor([1.2354e-04, 1.0516e-04, 9.3747e-05, 1.1948e-04, 1.0224e-04, 1.5327e-04, + 1.0743e-04, 1.1392e-04], device='cuda:1') +2023-02-07 10:35:17,344 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207567.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:35:20,581 INFO [train.py:901] (1/4) Epoch 26, batch 5500, loss[loss=0.1966, simple_loss=0.2741, pruned_loss=0.0595, over 7969.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.284, pruned_loss=0.0584, over 1609985.33 frames. ], batch size: 21, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:35:34,396 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=207592.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:35:47,182 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.395e+02 2.975e+02 3.460e+02 7.775e+02, threshold=5.949e+02, percent-clipped=2.0 +2023-02-07 10:35:56,713 INFO [train.py:901] (1/4) Epoch 26, batch 5550, loss[loss=0.2029, simple_loss=0.2907, pruned_loss=0.05759, over 8361.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2838, pruned_loss=0.05798, over 1608000.20 frames. ], batch size: 24, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:36:17,884 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207653.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:36:30,422 INFO [train.py:901] (1/4) Epoch 26, batch 5600, loss[loss=0.1685, simple_loss=0.2597, pruned_loss=0.03864, over 8186.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2839, pruned_loss=0.05836, over 1608471.50 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:36:55,068 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.412e+02 3.079e+02 3.750e+02 8.490e+02, threshold=6.158e+02, percent-clipped=5.0 +2023-02-07 10:37:04,589 INFO [train.py:901] (1/4) Epoch 26, batch 5650, loss[loss=0.2457, simple_loss=0.3378, pruned_loss=0.07681, over 8202.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2841, pruned_loss=0.05823, over 1610818.18 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:37:12,978 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 10:37:26,295 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.35 vs. limit=5.0 +2023-02-07 10:37:40,859 INFO [train.py:901] (1/4) Epoch 26, batch 5700, loss[loss=0.1683, simple_loss=0.2514, pruned_loss=0.04263, over 7926.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2827, pruned_loss=0.05772, over 1610521.47 frames. ], batch size: 20, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:38:05,995 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.232e+02 2.912e+02 3.330e+02 6.698e+02, threshold=5.824e+02, percent-clipped=1.0 +2023-02-07 10:38:14,804 INFO [train.py:901] (1/4) Epoch 26, batch 5750, loss[loss=0.2032, simple_loss=0.2863, pruned_loss=0.06002, over 8438.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2833, pruned_loss=0.05789, over 1609117.93 frames. ], batch size: 27, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:38:16,900 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 10:38:50,531 INFO [train.py:901] (1/4) Epoch 26, batch 5800, loss[loss=0.1958, simple_loss=0.2858, pruned_loss=0.05291, over 8291.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.282, pruned_loss=0.05731, over 1608579.66 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:39:15,995 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.614e+02 3.148e+02 4.020e+02 8.026e+02, threshold=6.297e+02, percent-clipped=4.0 +2023-02-07 10:39:16,251 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207909.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:39:24,803 INFO [train.py:901] (1/4) Epoch 26, batch 5850, loss[loss=0.2251, simple_loss=0.3126, pruned_loss=0.06883, over 8320.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2822, pruned_loss=0.0576, over 1608026.54 frames. ], batch size: 25, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:39:32,993 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=207934.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:39:59,794 INFO [train.py:901] (1/4) Epoch 26, batch 5900, loss[loss=0.1659, simple_loss=0.2644, pruned_loss=0.03376, over 8282.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2833, pruned_loss=0.05821, over 1611775.29 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:40:15,813 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7197, 1.4682, 2.8377, 1.4835, 2.2023, 3.0768, 3.2035, 2.6281], + device='cuda:1'), covar=tensor([0.1134, 0.1537, 0.0371, 0.1911, 0.0856, 0.0293, 0.0574, 0.0606], + device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0324, 0.0290, 0.0316, 0.0317, 0.0275, 0.0435, 0.0305], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 10:40:26,852 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.451e+02 2.961e+02 3.656e+02 5.483e+02, threshold=5.923e+02, percent-clipped=0.0 +2023-02-07 10:40:35,637 INFO [train.py:901] (1/4) Epoch 26, batch 5950, loss[loss=0.1668, simple_loss=0.2437, pruned_loss=0.04495, over 7802.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2828, pruned_loss=0.05819, over 1614585.22 frames. ], batch size: 19, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:40:54,933 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208050.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:41:09,813 INFO [train.py:901] (1/4) Epoch 26, batch 6000, loss[loss=0.1875, simple_loss=0.2667, pruned_loss=0.0541, over 7805.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.283, pruned_loss=0.0583, over 1613381.12 frames. ], batch size: 20, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:41:09,813 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 10:41:24,448 INFO [train.py:935] (1/4) Epoch 26, validation: loss=0.1721, simple_loss=0.2717, pruned_loss=0.03627, over 944034.00 frames. +2023-02-07 10:41:24,449 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 10:41:32,221 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2333, 1.0656, 1.3213, 1.0426, 1.0010, 1.3274, 0.0729, 0.9647], + device='cuda:1'), covar=tensor([0.1370, 0.1331, 0.0507, 0.0682, 0.2228, 0.0555, 0.1842, 0.1069], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0203, 0.0133, 0.0222, 0.0276, 0.0144, 0.0170, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 10:41:51,019 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.308e+02 2.837e+02 3.630e+02 6.769e+02, threshold=5.675e+02, percent-clipped=2.0 +2023-02-07 10:42:00,859 INFO [train.py:901] (1/4) Epoch 26, batch 6050, loss[loss=0.186, simple_loss=0.2655, pruned_loss=0.05328, over 7800.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2825, pruned_loss=0.05805, over 1614514.44 frames. ], batch size: 20, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:42:25,845 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2208, 4.1948, 3.7969, 1.9999, 3.6420, 3.7843, 3.7046, 3.5815], + device='cuda:1'), covar=tensor([0.0658, 0.0528, 0.0954, 0.4114, 0.0867, 0.1008, 0.1260, 0.0789], + device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0446, 0.0435, 0.0541, 0.0427, 0.0451, 0.0427, 0.0395], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:42:36,036 INFO [train.py:901] (1/4) Epoch 26, batch 6100, loss[loss=0.2172, simple_loss=0.2952, pruned_loss=0.06955, over 8464.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2819, pruned_loss=0.05759, over 1615994.16 frames. ], batch size: 28, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:42:48,603 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 10:43:01,339 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.708e+02 2.422e+02 2.947e+02 3.994e+02 1.088e+03, threshold=5.894e+02, percent-clipped=8.0 +2023-02-07 10:43:06,118 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5508, 4.6323, 4.0911, 2.0302, 3.9774, 4.1655, 4.1113, 3.9577], + device='cuda:1'), covar=tensor([0.0717, 0.0505, 0.1109, 0.4645, 0.0897, 0.0967, 0.1194, 0.0813], + device='cuda:1'), in_proj_covar=tensor([0.0532, 0.0450, 0.0438, 0.0545, 0.0429, 0.0454, 0.0429, 0.0397], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:43:10,773 INFO [train.py:901] (1/4) Epoch 26, batch 6150, loss[loss=0.2279, simple_loss=0.3004, pruned_loss=0.07776, over 8100.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2831, pruned_loss=0.05849, over 1616410.38 frames. ], batch size: 23, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:43:45,589 INFO [train.py:901] (1/4) Epoch 26, batch 6200, loss[loss=0.2158, simple_loss=0.309, pruned_loss=0.06135, over 7972.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2841, pruned_loss=0.05906, over 1613394.67 frames. ], batch size: 21, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:43:53,058 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208283.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:44:10,211 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.196e+02 2.837e+02 3.308e+02 7.178e+02, threshold=5.674e+02, percent-clipped=2.0 +2023-02-07 10:44:18,995 INFO [train.py:901] (1/4) Epoch 26, batch 6250, loss[loss=0.2239, simple_loss=0.3099, pruned_loss=0.06892, over 8353.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2841, pruned_loss=0.0594, over 1614402.53 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:44:55,647 INFO [train.py:901] (1/4) Epoch 26, batch 6300, loss[loss=0.2211, simple_loss=0.3031, pruned_loss=0.06959, over 8360.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2836, pruned_loss=0.05942, over 1607988.29 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:45:10,476 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=208394.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:45:20,552 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.443e+02 2.919e+02 3.618e+02 1.192e+03, threshold=5.838e+02, percent-clipped=3.0 +2023-02-07 10:45:29,139 INFO [train.py:901] (1/4) Epoch 26, batch 6350, loss[loss=0.1782, simple_loss=0.2666, pruned_loss=0.04486, over 8470.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2831, pruned_loss=0.05898, over 1607360.98 frames. ], batch size: 25, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:46:04,931 INFO [train.py:901] (1/4) Epoch 26, batch 6400, loss[loss=0.1915, simple_loss=0.2741, pruned_loss=0.05446, over 8102.00 frames. ], tot_loss[loss=0.199, simple_loss=0.282, pruned_loss=0.05797, over 1605507.71 frames. ], batch size: 21, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:46:30,394 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1753, 2.0187, 2.5771, 2.1575, 2.5351, 2.3035, 2.1379, 1.4000], + device='cuda:1'), covar=tensor([0.5890, 0.4984, 0.2086, 0.4084, 0.2847, 0.3191, 0.2015, 0.5608], + device='cuda:1'), in_proj_covar=tensor([0.0962, 0.1013, 0.0831, 0.0985, 0.1022, 0.0923, 0.0770, 0.0847], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:46:30,799 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.580e+02 3.188e+02 3.813e+02 6.849e+02, threshold=6.376e+02, percent-clipped=3.0 +2023-02-07 10:46:31,000 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208509.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:46:39,660 INFO [train.py:901] (1/4) Epoch 26, batch 6450, loss[loss=0.2018, simple_loss=0.2858, pruned_loss=0.05894, over 7798.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2817, pruned_loss=0.05788, over 1608097.04 frames. ], batch size: 20, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:47:13,750 INFO [train.py:901] (1/4) Epoch 26, batch 6500, loss[loss=0.1792, simple_loss=0.2731, pruned_loss=0.04261, over 8104.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2821, pruned_loss=0.05782, over 1606461.55 frames. ], batch size: 23, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:47:29,486 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208594.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:47:39,232 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.371e+02 2.869e+02 3.528e+02 8.936e+02, threshold=5.738e+02, percent-clipped=3.0 +2023-02-07 10:47:44,302 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6075, 2.4053, 3.1162, 2.5570, 3.0939, 2.5440, 2.4283, 2.0950], + device='cuda:1'), covar=tensor([0.4949, 0.5044, 0.2114, 0.3856, 0.2424, 0.2944, 0.1842, 0.5121], + device='cuda:1'), in_proj_covar=tensor([0.0968, 0.1019, 0.0835, 0.0991, 0.1025, 0.0928, 0.0773, 0.0852], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:47:48,763 INFO [train.py:901] (1/4) Epoch 26, batch 6550, loss[loss=0.2173, simple_loss=0.294, pruned_loss=0.07029, over 7917.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2829, pruned_loss=0.05824, over 1608015.09 frames. ], batch size: 20, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:47:52,200 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=208627.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:47:56,158 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 10:48:12,882 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 10:48:22,503 INFO [train.py:901] (1/4) Epoch 26, batch 6600, loss[loss=0.1867, simple_loss=0.279, pruned_loss=0.04718, over 8344.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2837, pruned_loss=0.05847, over 1608527.38 frames. ], batch size: 24, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:48:49,203 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.581e+02 2.930e+02 3.571e+02 6.165e+02, threshold=5.859e+02, percent-clipped=2.0 +2023-02-07 10:48:58,727 INFO [train.py:901] (1/4) Epoch 26, batch 6650, loss[loss=0.1756, simple_loss=0.254, pruned_loss=0.04866, over 7788.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2831, pruned_loss=0.05848, over 1608080.07 frames. ], batch size: 19, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:49:12,771 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208742.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:49:28,886 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208765.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:49:29,622 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0849, 1.9444, 2.5046, 2.0997, 2.5845, 2.1675, 1.9448, 1.4302], + device='cuda:1'), covar=tensor([0.6398, 0.5282, 0.2119, 0.4288, 0.2875, 0.3535, 0.2346, 0.5742], + device='cuda:1'), in_proj_covar=tensor([0.0967, 0.1017, 0.0834, 0.0989, 0.1024, 0.0927, 0.0771, 0.0849], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:49:33,544 INFO [train.py:901] (1/4) Epoch 26, batch 6700, loss[loss=0.2676, simple_loss=0.3323, pruned_loss=0.1015, over 8471.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2837, pruned_loss=0.05887, over 1611136.86 frames. ], batch size: 25, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:49:46,224 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208790.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:49:59,643 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 2.530e+02 3.053e+02 4.076e+02 9.744e+02, threshold=6.106e+02, percent-clipped=7.0 +2023-02-07 10:50:03,310 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8437, 1.8888, 1.7553, 2.3139, 1.0206, 1.5797, 1.7372, 1.8598], + device='cuda:1'), covar=tensor([0.0729, 0.0715, 0.0897, 0.0389, 0.1113, 0.1218, 0.0720, 0.0804], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0195, 0.0245, 0.0212, 0.0203, 0.0246, 0.0250, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 10:50:09,371 INFO [train.py:901] (1/4) Epoch 26, batch 6750, loss[loss=0.1871, simple_loss=0.2608, pruned_loss=0.05667, over 7447.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2828, pruned_loss=0.05842, over 1610375.52 frames. ], batch size: 17, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:50:30,943 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 10:50:43,653 INFO [train.py:901] (1/4) Epoch 26, batch 6800, loss[loss=0.1868, simple_loss=0.2601, pruned_loss=0.05677, over 7539.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2842, pruned_loss=0.05885, over 1612829.60 frames. ], batch size: 18, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:50:57,054 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-07 10:51:08,807 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.376e+02 2.847e+02 3.449e+02 1.016e+03, threshold=5.694e+02, percent-clipped=2.0 +2023-02-07 10:51:18,814 INFO [train.py:901] (1/4) Epoch 26, batch 6850, loss[loss=0.2016, simple_loss=0.2923, pruned_loss=0.05546, over 8465.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2853, pruned_loss=0.05926, over 1619025.58 frames. ], batch size: 25, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:51:19,504 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 10:51:30,533 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=208938.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:51:33,384 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9110, 1.9233, 1.7128, 2.5562, 1.1556, 1.5706, 1.9141, 1.9919], + device='cuda:1'), covar=tensor([0.0730, 0.0870, 0.0916, 0.0396, 0.1124, 0.1293, 0.0848, 0.0819], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0194, 0.0244, 0.0212, 0.0202, 0.0245, 0.0249, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 10:51:37,663 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3975, 2.1801, 2.6987, 2.2583, 2.7936, 2.4549, 2.2919, 1.6994], + device='cuda:1'), covar=tensor([0.5631, 0.5141, 0.2238, 0.4033, 0.2642, 0.3071, 0.1877, 0.5387], + device='cuda:1'), in_proj_covar=tensor([0.0961, 0.1014, 0.0830, 0.0986, 0.1020, 0.0924, 0.0769, 0.0847], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:51:42,402 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5376, 2.3314, 3.1221, 2.4592, 3.0285, 2.5789, 2.4390, 1.8379], + device='cuda:1'), covar=tensor([0.5550, 0.5400, 0.2074, 0.4045, 0.2640, 0.3141, 0.2065, 0.5805], + device='cuda:1'), in_proj_covar=tensor([0.0961, 0.1013, 0.0829, 0.0986, 0.1019, 0.0924, 0.0769, 0.0846], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:51:45,310 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-07 10:51:54,559 INFO [train.py:901] (1/4) Epoch 26, batch 6900, loss[loss=0.1849, simple_loss=0.2492, pruned_loss=0.06036, over 7711.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2844, pruned_loss=0.05907, over 1615232.22 frames. ], batch size: 18, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:52:10,630 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0527, 3.4645, 2.0649, 2.7825, 2.5807, 2.0380, 2.5756, 2.8955], + device='cuda:1'), covar=tensor([0.1724, 0.0399, 0.1314, 0.0818, 0.0846, 0.1440, 0.1163, 0.1127], + device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0239, 0.0339, 0.0310, 0.0302, 0.0345, 0.0347, 0.0322], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 10:52:12,020 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208998.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:52:12,736 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9753, 1.7718, 2.0643, 1.8473, 2.0352, 2.0129, 1.8743, 0.8468], + device='cuda:1'), covar=tensor([0.6423, 0.4972, 0.2283, 0.4033, 0.2714, 0.3428, 0.2224, 0.5785], + device='cuda:1'), in_proj_covar=tensor([0.0961, 0.1013, 0.0829, 0.0986, 0.1018, 0.0923, 0.0769, 0.0846], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:52:20,225 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.430e+02 2.933e+02 3.890e+02 9.541e+02, threshold=5.866e+02, percent-clipped=7.0 +2023-02-07 10:52:23,035 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 10:52:23,719 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4530, 4.4044, 3.9985, 2.5645, 3.8605, 4.0393, 4.0389, 3.8535], + device='cuda:1'), covar=tensor([0.0641, 0.0578, 0.1051, 0.3713, 0.0895, 0.1075, 0.1138, 0.0873], + device='cuda:1'), in_proj_covar=tensor([0.0537, 0.0456, 0.0442, 0.0552, 0.0436, 0.0460, 0.0436, 0.0402], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:52:28,982 INFO [train.py:901] (1/4) Epoch 26, batch 6950, loss[loss=0.1895, simple_loss=0.2589, pruned_loss=0.06006, over 7455.00 frames. ], tot_loss[loss=0.201, simple_loss=0.284, pruned_loss=0.05897, over 1615573.84 frames. ], batch size: 17, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:52:29,848 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209023.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:52:51,124 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209053.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:52:51,873 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8970, 1.7897, 2.8074, 2.2059, 2.6456, 1.9454, 1.7262, 1.4495], + device='cuda:1'), covar=tensor([0.7524, 0.6713, 0.2244, 0.4677, 0.3197, 0.4711, 0.3053, 0.6209], + device='cuda:1'), in_proj_covar=tensor([0.0961, 0.1013, 0.0829, 0.0986, 0.1017, 0.0922, 0.0770, 0.0847], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 10:53:04,217 INFO [train.py:901] (1/4) Epoch 26, batch 7000, loss[loss=0.2117, simple_loss=0.296, pruned_loss=0.06371, over 8512.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.284, pruned_loss=0.05883, over 1616263.50 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:53:30,442 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.543e+02 3.075e+02 4.223e+02 1.225e+03, threshold=6.150e+02, percent-clipped=4.0 +2023-02-07 10:53:38,407 INFO [train.py:901] (1/4) Epoch 26, batch 7050, loss[loss=0.203, simple_loss=0.266, pruned_loss=0.07005, over 7797.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2839, pruned_loss=0.05917, over 1610734.18 frames. ], batch size: 19, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:53:44,279 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7329, 2.0906, 3.1409, 1.6360, 2.5497, 2.1796, 1.7710, 2.5156], + device='cuda:1'), covar=tensor([0.1861, 0.2679, 0.0950, 0.4457, 0.1816, 0.3141, 0.2405, 0.2289], + device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0633, 0.0563, 0.0668, 0.0660, 0.0610, 0.0561, 0.0648], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:54:15,165 INFO [train.py:901] (1/4) Epoch 26, batch 7100, loss[loss=0.1685, simple_loss=0.2443, pruned_loss=0.04634, over 7553.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2828, pruned_loss=0.05878, over 1609311.35 frames. ], batch size: 18, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:54:42,202 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.719e+02 2.538e+02 3.057e+02 3.964e+02 1.199e+03, threshold=6.114e+02, percent-clipped=9.0 +2023-02-07 10:54:49,195 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209220.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:54:50,427 INFO [train.py:901] (1/4) Epoch 26, batch 7150, loss[loss=0.2141, simple_loss=0.3027, pruned_loss=0.0628, over 8037.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2832, pruned_loss=0.0593, over 1607083.87 frames. ], batch size: 22, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:55:25,278 INFO [train.py:901] (1/4) Epoch 26, batch 7200, loss[loss=0.2048, simple_loss=0.2869, pruned_loss=0.06139, over 8517.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2833, pruned_loss=0.05893, over 1613096.68 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:55:26,110 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209273.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:55:51,833 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209309.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:55:52,284 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.329e+02 2.733e+02 3.562e+02 6.414e+02, threshold=5.467e+02, percent-clipped=2.0 +2023-02-07 10:55:56,509 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209316.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:56:00,335 INFO [train.py:901] (1/4) Epoch 26, batch 7250, loss[loss=0.1632, simple_loss=0.2426, pruned_loss=0.04192, over 7223.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2832, pruned_loss=0.0588, over 1612984.47 frames. ], batch size: 16, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:56:08,438 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209334.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:56:33,562 INFO [train.py:901] (1/4) Epoch 26, batch 7300, loss[loss=0.1827, simple_loss=0.268, pruned_loss=0.04868, over 8325.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2835, pruned_loss=0.05863, over 1616375.34 frames. ], batch size: 25, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:56:56,000 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.2586, 4.2311, 3.8897, 1.9756, 3.8182, 3.8104, 3.7845, 3.6900], + device='cuda:1'), covar=tensor([0.0714, 0.0525, 0.1027, 0.4299, 0.0824, 0.1010, 0.1233, 0.0913], + device='cuda:1'), in_proj_covar=tensor([0.0541, 0.0458, 0.0445, 0.0556, 0.0439, 0.0464, 0.0439, 0.0404], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 10:56:59,846 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.697e+02 2.463e+02 3.055e+02 3.934e+02 7.151e+02, threshold=6.111e+02, percent-clipped=5.0 +2023-02-07 10:57:06,738 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 10:57:09,437 INFO [train.py:901] (1/4) Epoch 26, batch 7350, loss[loss=0.2221, simple_loss=0.2958, pruned_loss=0.07418, over 8083.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2842, pruned_loss=0.05929, over 1612567.49 frames. ], batch size: 21, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:57:26,304 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 10:57:42,849 INFO [train.py:901] (1/4) Epoch 26, batch 7400, loss[loss=0.2055, simple_loss=0.2873, pruned_loss=0.06182, over 8350.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2854, pruned_loss=0.06003, over 1613550.14 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:57:50,999 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209484.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:58:04,813 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 10:58:09,479 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.534e+02 3.042e+02 3.812e+02 9.347e+02, threshold=6.084e+02, percent-clipped=5.0 +2023-02-07 10:58:17,131 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8496, 2.2568, 3.7000, 1.8778, 1.7562, 3.5287, 0.6879, 2.1392], + device='cuda:1'), covar=tensor([0.1481, 0.1434, 0.0317, 0.1823, 0.2645, 0.0410, 0.2226, 0.1555], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0205, 0.0133, 0.0224, 0.0276, 0.0144, 0.0172, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 10:58:17,541 INFO [train.py:901] (1/4) Epoch 26, batch 7450, loss[loss=0.185, simple_loss=0.2662, pruned_loss=0.05194, over 7419.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2854, pruned_loss=0.0599, over 1606527.88 frames. ], batch size: 17, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:58:22,478 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209528.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:58:47,408 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209564.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:58:52,736 INFO [train.py:901] (1/4) Epoch 26, batch 7500, loss[loss=0.1789, simple_loss=0.2498, pruned_loss=0.05396, over 7529.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2847, pruned_loss=0.0596, over 1605715.64 frames. ], batch size: 18, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:59:18,686 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.436e+02 2.983e+02 3.503e+02 8.056e+02, threshold=5.967e+02, percent-clipped=5.0 +2023-02-07 10:59:24,168 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209617.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:59:27,346 INFO [train.py:901] (1/4) Epoch 26, batch 7550, loss[loss=0.1783, simple_loss=0.269, pruned_loss=0.04378, over 8293.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2839, pruned_loss=0.05925, over 1608262.44 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 10:59:28,767 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209624.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:59:54,106 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209660.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:59:55,495 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2403, 3.1803, 2.9535, 1.6273, 2.8603, 2.9392, 2.8567, 2.8045], + device='cuda:1'), covar=tensor([0.1139, 0.0830, 0.1347, 0.4468, 0.1067, 0.1423, 0.1495, 0.1149], + device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0456, 0.0443, 0.0554, 0.0436, 0.0462, 0.0436, 0.0403], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:00:01,950 INFO [train.py:901] (1/4) Epoch 26, batch 7600, loss[loss=0.1891, simple_loss=0.2757, pruned_loss=0.05119, over 8322.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2828, pruned_loss=0.05909, over 1601499.20 frames. ], batch size: 25, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:00:07,023 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209679.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:00:25,098 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209706.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:00:27,679 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.404e+02 2.880e+02 3.478e+02 6.437e+02, threshold=5.761e+02, percent-clipped=3.0 +2023-02-07 11:00:35,718 INFO [train.py:901] (1/4) Epoch 26, batch 7650, loss[loss=0.2215, simple_loss=0.3012, pruned_loss=0.07088, over 8684.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2819, pruned_loss=0.05806, over 1607632.55 frames. ], batch size: 34, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:00:43,081 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209732.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:01:00,268 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.61 vs. limit=5.0 +2023-02-07 11:01:10,731 INFO [train.py:901] (1/4) Epoch 26, batch 7700, loss[loss=0.1898, simple_loss=0.2819, pruned_loss=0.04889, over 8195.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2822, pruned_loss=0.05833, over 1606301.39 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:01:12,901 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209775.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:01:14,130 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 11:01:23,009 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 +2023-02-07 11:01:36,858 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 2.688e+02 3.083e+02 3.850e+02 9.382e+02, threshold=6.167e+02, percent-clipped=8.0 +2023-02-07 11:01:39,183 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6000, 2.0644, 3.2880, 1.4981, 2.4906, 2.0346, 1.6859, 2.4718], + device='cuda:1'), covar=tensor([0.2008, 0.2720, 0.0909, 0.4843, 0.1969, 0.3493, 0.2601, 0.2435], + device='cuda:1'), in_proj_covar=tensor([0.0537, 0.0631, 0.0560, 0.0666, 0.0659, 0.0610, 0.0559, 0.0644], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:01:44,902 INFO [train.py:901] (1/4) Epoch 26, batch 7750, loss[loss=0.209, simple_loss=0.298, pruned_loss=0.05999, over 8328.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2828, pruned_loss=0.05863, over 1608964.27 frames. ], batch size: 26, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:01:48,919 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209828.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:02:19,759 INFO [train.py:901] (1/4) Epoch 26, batch 7800, loss[loss=0.1945, simple_loss=0.2881, pruned_loss=0.05048, over 8461.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2829, pruned_loss=0.05883, over 1606537.07 frames. ], batch size: 25, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:02:19,817 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209872.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:02:44,108 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2811, 1.3473, 3.4028, 1.1706, 3.0096, 2.9091, 3.1221, 3.0477], + device='cuda:1'), covar=tensor([0.0897, 0.4483, 0.0924, 0.4144, 0.1452, 0.1151, 0.0830, 0.0919], + device='cuda:1'), in_proj_covar=tensor([0.0672, 0.0666, 0.0734, 0.0658, 0.0742, 0.0633, 0.0633, 0.0709], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:02:45,323 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.465e+02 2.962e+02 3.430e+02 5.705e+02, threshold=5.924e+02, percent-clipped=0.0 +2023-02-07 11:02:53,239 INFO [train.py:901] (1/4) Epoch 26, batch 7850, loss[loss=0.1639, simple_loss=0.2476, pruned_loss=0.0401, over 8092.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2842, pruned_loss=0.05921, over 1608413.26 frames. ], batch size: 21, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:03:01,889 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209935.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:07,204 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209943.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:18,181 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209960.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:23,279 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209968.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:25,900 INFO [train.py:901] (1/4) Epoch 26, batch 7900, loss[loss=0.1907, simple_loss=0.2666, pruned_loss=0.05747, over 7654.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2841, pruned_loss=0.05912, over 1613093.69 frames. ], batch size: 19, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:03:35,883 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209987.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:36,568 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209988.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:51,904 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.305e+02 2.795e+02 3.387e+02 5.942e+02, threshold=5.591e+02, percent-clipped=1.0 +2023-02-07 11:03:54,075 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210013.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:59,749 INFO [train.py:901] (1/4) Epoch 26, batch 7950, loss[loss=0.2109, simple_loss=0.2933, pruned_loss=0.06425, over 8257.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2852, pruned_loss=0.0599, over 1612811.64 frames. ], batch size: 24, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:04:06,009 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210031.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:18,572 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=210050.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:22,765 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210056.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:33,086 INFO [train.py:901] (1/4) Epoch 26, batch 8000, loss[loss=0.1752, simple_loss=0.2586, pruned_loss=0.04586, over 7805.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2836, pruned_loss=0.05929, over 1605458.49 frames. ], batch size: 20, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:04:35,217 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210075.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:40,668 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210083.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:58,002 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.316e+02 2.819e+02 3.710e+02 9.270e+02, threshold=5.638e+02, percent-clipped=7.0 +2023-02-07 11:05:05,834 INFO [train.py:901] (1/4) Epoch 26, batch 8050, loss[loss=0.2088, simple_loss=0.2826, pruned_loss=0.0675, over 7921.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2804, pruned_loss=0.0584, over 1585281.69 frames. ], batch size: 20, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:05:38,157 WARNING [train.py:1067] (1/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 11:05:43,121 INFO [train.py:901] (1/4) Epoch 27, batch 0, loss[loss=0.1783, simple_loss=0.2504, pruned_loss=0.05316, over 7252.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2504, pruned_loss=0.05316, over 7252.00 frames. ], batch size: 16, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:05:43,121 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 11:05:54,193 INFO [train.py:935] (1/4) Epoch 27, validation: loss=0.172, simple_loss=0.2713, pruned_loss=0.03628, over 944034.00 frames. +2023-02-07 11:05:54,195 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 11:06:01,165 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210165.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:06:08,368 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 11:06:24,777 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210199.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:06:28,647 INFO [train.py:901] (1/4) Epoch 27, batch 50, loss[loss=0.1946, simple_loss=0.265, pruned_loss=0.06205, over 7539.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2828, pruned_loss=0.05972, over 361780.31 frames. ], batch size: 18, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:06:33,573 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.417e+02 2.930e+02 3.516e+02 7.088e+02, threshold=5.860e+02, percent-clipped=5.0 +2023-02-07 11:06:41,899 WARNING [train.py:1067] (1/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 11:06:43,309 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210224.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:06:56,868 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210243.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:07:04,542 INFO [train.py:901] (1/4) Epoch 27, batch 100, loss[loss=0.2043, simple_loss=0.2838, pruned_loss=0.06239, over 7639.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2826, pruned_loss=0.05891, over 637334.43 frames. ], batch size: 19, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:07:05,184 WARNING [train.py:1067] (1/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 11:07:13,381 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210268.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:07:37,991 INFO [train.py:901] (1/4) Epoch 27, batch 150, loss[loss=0.1725, simple_loss=0.2415, pruned_loss=0.05173, over 4945.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2844, pruned_loss=0.05922, over 855270.38 frames. ], batch size: 11, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:07:41,158 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.352e+02 2.905e+02 3.661e+02 1.089e+03, threshold=5.811e+02, percent-clipped=3.0 +2023-02-07 11:07:41,641 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 +2023-02-07 11:08:02,857 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210339.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:08:14,086 INFO [train.py:901] (1/4) Epoch 27, batch 200, loss[loss=0.2122, simple_loss=0.2921, pruned_loss=0.06613, over 8605.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2842, pruned_loss=0.059, over 1024424.00 frames. ], batch size: 34, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:08:20,492 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210364.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:08:48,349 INFO [train.py:901] (1/4) Epoch 27, batch 250, loss[loss=0.1786, simple_loss=0.2668, pruned_loss=0.04518, over 8254.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2838, pruned_loss=0.05854, over 1157542.17 frames. ], batch size: 24, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:08:49,370 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 11:08:51,589 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.304e+02 2.819e+02 3.559e+02 6.263e+02, threshold=5.638e+02, percent-clipped=1.0 +2023-02-07 11:08:57,573 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 11:08:57,625 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=210419.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:08:59,061 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210421.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:09:06,476 WARNING [train.py:1067] (1/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 11:09:10,762 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1579, 2.0223, 2.6738, 2.2558, 2.7452, 2.2545, 2.1528, 1.5413], + device='cuda:1'), covar=tensor([0.6253, 0.5226, 0.2212, 0.4176, 0.2731, 0.3337, 0.2040, 0.5844], + device='cuda:1'), in_proj_covar=tensor([0.0965, 0.1017, 0.0828, 0.0985, 0.1021, 0.0926, 0.0772, 0.0847], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 11:09:15,930 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210446.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:09:23,015 INFO [train.py:901] (1/4) Epoch 27, batch 300, loss[loss=0.1779, simple_loss=0.2614, pruned_loss=0.04715, over 7917.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.285, pruned_loss=0.05879, over 1265464.35 frames. ], batch size: 20, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:09:57,348 INFO [train.py:901] (1/4) Epoch 27, batch 350, loss[loss=0.1838, simple_loss=0.2793, pruned_loss=0.04409, over 8080.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2832, pruned_loss=0.05729, over 1340474.10 frames. ], batch size: 21, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:10:00,691 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.321e+02 2.740e+02 3.479e+02 7.751e+02, threshold=5.481e+02, percent-clipped=4.0 +2023-02-07 11:10:06,929 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4301, 1.7904, 1.3686, 3.0587, 1.4193, 1.4694, 2.2034, 2.0165], + device='cuda:1'), covar=tensor([0.1680, 0.1380, 0.2011, 0.0326, 0.1335, 0.1839, 0.0929, 0.1044], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0194, 0.0246, 0.0211, 0.0202, 0.0245, 0.0249, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 11:10:16,900 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210534.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:10:29,881 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 11:10:30,766 INFO [train.py:901] (1/4) Epoch 27, batch 400, loss[loss=0.1986, simple_loss=0.2914, pruned_loss=0.05286, over 8501.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2832, pruned_loss=0.05745, over 1404708.88 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:11:06,891 INFO [train.py:901] (1/4) Epoch 27, batch 450, loss[loss=0.2037, simple_loss=0.2937, pruned_loss=0.05689, over 8467.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.283, pruned_loss=0.05775, over 1453086.73 frames. ], batch size: 25, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:11:10,233 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 2.445e+02 3.096e+02 3.744e+02 6.670e+02, threshold=6.192e+02, percent-clipped=5.0 +2023-02-07 11:11:28,567 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.4493, 4.3997, 3.9842, 2.3056, 3.8543, 4.0768, 3.9574, 3.9322], + device='cuda:1'), covar=tensor([0.0681, 0.0558, 0.0963, 0.4098, 0.0819, 0.1014, 0.1226, 0.0819], + device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0458, 0.0442, 0.0555, 0.0437, 0.0460, 0.0436, 0.0405], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:11:37,202 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210650.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:11:40,169 INFO [train.py:901] (1/4) Epoch 27, batch 500, loss[loss=0.2368, simple_loss=0.3186, pruned_loss=0.07749, over 8567.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.284, pruned_loss=0.05845, over 1492111.16 frames. ], batch size: 31, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:11:55,130 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4628, 2.7272, 3.0002, 1.8524, 3.1594, 2.0270, 1.6633, 2.4617], + device='cuda:1'), covar=tensor([0.0819, 0.0458, 0.0405, 0.0851, 0.0584, 0.0943, 0.0997, 0.0559], + device='cuda:1'), in_proj_covar=tensor([0.0469, 0.0406, 0.0360, 0.0458, 0.0392, 0.0548, 0.0402, 0.0436], + device='cuda:1'), out_proj_covar=tensor([1.2454e-04, 1.0560e-04, 9.4168e-05, 1.2007e-04, 1.0262e-04, 1.5318e-04, + 1.0753e-04, 1.1439e-04], device='cuda:1') +2023-02-07 11:12:01,202 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210684.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:12:12,486 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 11:12:12,810 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210700.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:12:15,969 INFO [train.py:901] (1/4) Epoch 27, batch 550, loss[loss=0.2183, simple_loss=0.2939, pruned_loss=0.07139, over 8131.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2834, pruned_loss=0.05822, over 1521009.90 frames. ], batch size: 22, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:12:19,365 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.336e+02 2.792e+02 3.793e+02 8.487e+02, threshold=5.584e+02, percent-clipped=3.0 +2023-02-07 11:12:50,315 INFO [train.py:901] (1/4) Epoch 27, batch 600, loss[loss=0.2284, simple_loss=0.3074, pruned_loss=0.0747, over 8467.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2833, pruned_loss=0.05828, over 1543230.11 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:13:08,465 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210782.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:13:11,587 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 11:13:13,656 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210790.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:13:23,584 INFO [train.py:901] (1/4) Epoch 27, batch 650, loss[loss=0.2271, simple_loss=0.3086, pruned_loss=0.07285, over 8582.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.283, pruned_loss=0.05808, over 1559936.53 frames. ], batch size: 39, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:13:28,267 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.351e+02 2.894e+02 3.474e+02 6.032e+02, threshold=5.788e+02, percent-clipped=3.0 +2023-02-07 11:13:32,521 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210815.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:13:59,775 INFO [train.py:901] (1/4) Epoch 27, batch 700, loss[loss=0.1541, simple_loss=0.2412, pruned_loss=0.03347, over 7701.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2821, pruned_loss=0.05749, over 1571727.84 frames. ], batch size: 18, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:14:28,999 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9118, 3.3060, 1.8508, 2.7104, 2.5257, 1.6123, 2.3609, 2.9757], + device='cuda:1'), covar=tensor([0.1857, 0.0552, 0.1612, 0.0795, 0.1048, 0.2157, 0.1460, 0.1025], + device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0239, 0.0338, 0.0309, 0.0301, 0.0343, 0.0345, 0.0320], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 11:14:32,720 INFO [train.py:901] (1/4) Epoch 27, batch 750, loss[loss=0.1846, simple_loss=0.2597, pruned_loss=0.05472, over 7202.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2824, pruned_loss=0.05824, over 1581394.37 frames. ], batch size: 16, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:14:35,973 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.536e+02 2.996e+02 3.960e+02 1.304e+03, threshold=5.993e+02, percent-clipped=7.0 +2023-02-07 11:14:54,973 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 11:15:04,237 WARNING [train.py:1067] (1/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 11:15:08,850 INFO [train.py:901] (1/4) Epoch 27, batch 800, loss[loss=0.1757, simple_loss=0.2647, pruned_loss=0.04337, over 8244.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2827, pruned_loss=0.05798, over 1592450.72 frames. ], batch size: 22, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:15:11,731 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.89 vs. limit=5.0 +2023-02-07 11:15:34,998 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=210994.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:15:38,617 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-07 11:15:40,388 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211002.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:15:42,228 INFO [train.py:901] (1/4) Epoch 27, batch 850, loss[loss=0.1916, simple_loss=0.286, pruned_loss=0.0486, over 7802.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2828, pruned_loss=0.05766, over 1600529.76 frames. ], batch size: 20, lr: 2.81e-03, grad_scale: 16.0 +2023-02-07 11:15:45,636 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.265e+02 2.725e+02 3.482e+02 8.151e+02, threshold=5.450e+02, percent-clipped=2.0 +2023-02-07 11:15:46,884 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 11:15:57,715 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211028.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:16:10,423 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211044.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:16:17,676 INFO [train.py:901] (1/4) Epoch 27, batch 900, loss[loss=0.2123, simple_loss=0.2959, pruned_loss=0.06434, over 8247.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2846, pruned_loss=0.05884, over 1602070.04 frames. ], batch size: 24, lr: 2.81e-03, grad_scale: 16.0 +2023-02-07 11:16:51,933 INFO [train.py:901] (1/4) Epoch 27, batch 950, loss[loss=0.1886, simple_loss=0.2744, pruned_loss=0.05141, over 8512.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2835, pruned_loss=0.05804, over 1606086.27 frames. ], batch size: 26, lr: 2.81e-03, grad_scale: 16.0 +2023-02-07 11:16:54,833 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211109.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:16:55,251 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.483e+02 2.981e+02 4.008e+02 9.530e+02, threshold=5.961e+02, percent-clipped=10.0 +2023-02-07 11:17:06,151 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211126.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:17:11,860 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-07 11:17:17,580 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211143.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:17:18,070 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 11:17:26,828 INFO [train.py:901] (1/4) Epoch 27, batch 1000, loss[loss=0.2429, simple_loss=0.3287, pruned_loss=0.07855, over 8739.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2829, pruned_loss=0.05771, over 1606335.07 frames. ], batch size: 30, lr: 2.81e-03, grad_scale: 16.0 +2023-02-07 11:17:30,481 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211159.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:17:51,438 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([0.9938, 1.6110, 1.3310, 1.6091, 1.2824, 1.1556, 1.3236, 1.3020], + device='cuda:1'), covar=tensor([0.1259, 0.0543, 0.1478, 0.0572, 0.0975, 0.1773, 0.0975, 0.0823], + device='cuda:1'), in_proj_covar=tensor([0.0360, 0.0239, 0.0340, 0.0311, 0.0304, 0.0344, 0.0346, 0.0321], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 11:17:53,343 WARNING [train.py:1067] (1/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 11:18:03,293 INFO [train.py:901] (1/4) Epoch 27, batch 1050, loss[loss=0.2108, simple_loss=0.274, pruned_loss=0.07377, over 7724.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2822, pruned_loss=0.05745, over 1606600.26 frames. ], batch size: 18, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:18:05,225 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 11:18:07,260 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.750e+02 2.494e+02 3.070e+02 3.818e+02 8.233e+02, threshold=6.140e+02, percent-clipped=4.0 +2023-02-07 11:18:08,732 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3475, 2.0777, 2.5963, 2.2028, 2.5621, 2.3617, 2.2343, 1.5791], + device='cuda:1'), covar=tensor([0.5580, 0.4947, 0.2272, 0.4269, 0.2876, 0.3346, 0.1920, 0.5694], + device='cuda:1'), in_proj_covar=tensor([0.0964, 0.1018, 0.0831, 0.0988, 0.1021, 0.0928, 0.0771, 0.0850], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 11:18:27,160 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211241.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:18:36,463 INFO [train.py:901] (1/4) Epoch 27, batch 1100, loss[loss=0.211, simple_loss=0.2848, pruned_loss=0.06857, over 7795.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2825, pruned_loss=0.05769, over 1607283.16 frames. ], batch size: 19, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:19:13,344 INFO [train.py:901] (1/4) Epoch 27, batch 1150, loss[loss=0.1662, simple_loss=0.251, pruned_loss=0.0407, over 7782.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2828, pruned_loss=0.05774, over 1609802.37 frames. ], batch size: 19, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:19:15,965 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 11:19:17,147 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-07 11:19:17,280 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.371e+02 2.782e+02 3.549e+02 6.262e+02, threshold=5.564e+02, percent-clipped=1.0 +2023-02-07 11:19:40,860 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211346.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:19:46,762 INFO [train.py:901] (1/4) Epoch 27, batch 1200, loss[loss=0.21, simple_loss=0.3023, pruned_loss=0.0589, over 8440.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2836, pruned_loss=0.05872, over 1609957.26 frames. ], batch size: 27, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:19:53,750 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211365.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:08,485 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0970, 2.2709, 1.8280, 2.8380, 1.3948, 1.6356, 2.1063, 2.1846], + device='cuda:1'), covar=tensor([0.0681, 0.0777, 0.0831, 0.0338, 0.1085, 0.1281, 0.0800, 0.0779], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0194, 0.0244, 0.0211, 0.0202, 0.0244, 0.0249, 0.0204], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 11:20:11,156 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211390.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:18,556 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211399.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:21,056 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.6722, 4.7330, 4.2158, 2.1126, 4.1231, 4.3632, 4.3068, 4.1413], + device='cuda:1'), covar=tensor([0.0725, 0.0495, 0.1084, 0.4654, 0.0885, 0.0956, 0.1188, 0.0734], + device='cuda:1'), in_proj_covar=tensor([0.0543, 0.0459, 0.0446, 0.0557, 0.0440, 0.0463, 0.0437, 0.0407], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:20:21,765 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211404.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:22,040 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 11:20:22,323 INFO [train.py:901] (1/4) Epoch 27, batch 1250, loss[loss=0.1963, simple_loss=0.2914, pruned_loss=0.05054, over 8191.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2837, pruned_loss=0.0585, over 1613738.26 frames. ], batch size: 23, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:20:26,159 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 2.344e+02 2.922e+02 3.484e+02 6.390e+02, threshold=5.843e+02, percent-clipped=2.0 +2023-02-07 11:20:29,708 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211415.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:35,518 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211424.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:39,475 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211430.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:46,223 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211440.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:49,955 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-07 11:20:56,233 INFO [train.py:901] (1/4) Epoch 27, batch 1300, loss[loss=0.1446, simple_loss=0.2241, pruned_loss=0.03257, over 7435.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2825, pruned_loss=0.05798, over 1609807.84 frames. ], batch size: 17, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:20:59,839 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6022, 2.1120, 3.2898, 1.4775, 2.3251, 2.0223, 1.7564, 2.5060], + device='cuda:1'), covar=tensor([0.1871, 0.2453, 0.0973, 0.4679, 0.2032, 0.3248, 0.2311, 0.2268], + device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0632, 0.0564, 0.0669, 0.0659, 0.0610, 0.0560, 0.0644], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:21:00,455 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211461.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:21:23,172 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9700, 1.6807, 3.3117, 1.4349, 2.3113, 3.6074, 3.7316, 3.0919], + device='cuda:1'), covar=tensor([0.1235, 0.1704, 0.0307, 0.2191, 0.0946, 0.0246, 0.0537, 0.0522], + device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0327, 0.0293, 0.0321, 0.0320, 0.0278, 0.0437, 0.0307], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 11:21:24,574 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211497.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:21:30,516 INFO [train.py:901] (1/4) Epoch 27, batch 1350, loss[loss=0.1742, simple_loss=0.2522, pruned_loss=0.04809, over 7660.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2812, pruned_loss=0.05687, over 1610131.64 frames. ], batch size: 19, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:21:34,485 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.433e+02 2.859e+02 3.519e+02 6.900e+02, threshold=5.717e+02, percent-clipped=5.0 +2023-02-07 11:21:43,599 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211522.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:22:05,946 INFO [train.py:901] (1/4) Epoch 27, batch 1400, loss[loss=0.2124, simple_loss=0.285, pruned_loss=0.06994, over 8233.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2815, pruned_loss=0.05698, over 1608594.06 frames. ], batch size: 22, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:22:39,557 INFO [train.py:901] (1/4) Epoch 27, batch 1450, loss[loss=0.1933, simple_loss=0.2648, pruned_loss=0.0609, over 7212.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2814, pruned_loss=0.05704, over 1611086.15 frames. ], batch size: 16, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:22:43,650 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.739e+02 3.417e+02 5.363e+02 1.739e+03, threshold=6.835e+02, percent-clipped=22.0 +2023-02-07 11:22:45,702 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 11:23:15,882 INFO [train.py:901] (1/4) Epoch 27, batch 1500, loss[loss=0.1803, simple_loss=0.255, pruned_loss=0.05275, over 7437.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2811, pruned_loss=0.05705, over 1605362.17 frames. ], batch size: 17, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:23:42,394 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6053, 1.9186, 5.8018, 2.5943, 4.9093, 4.8054, 5.3844, 5.3670], + device='cuda:1'), covar=tensor([0.1265, 0.6937, 0.0712, 0.4535, 0.1961, 0.1562, 0.1118, 0.0817], + device='cuda:1'), in_proj_covar=tensor([0.0673, 0.0667, 0.0734, 0.0656, 0.0744, 0.0633, 0.0633, 0.0713], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:23:49,659 INFO [train.py:901] (1/4) Epoch 27, batch 1550, loss[loss=0.1848, simple_loss=0.2547, pruned_loss=0.05744, over 7436.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2808, pruned_loss=0.05712, over 1602697.05 frames. ], batch size: 17, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:23:53,681 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.372e+02 3.027e+02 3.476e+02 5.786e+02, threshold=6.054e+02, percent-clipped=0.0 +2023-02-07 11:23:57,834 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211717.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:24:15,283 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211742.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:24:19,434 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211748.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:24:24,775 INFO [train.py:901] (1/4) Epoch 27, batch 1600, loss[loss=0.2445, simple_loss=0.3181, pruned_loss=0.08546, over 8529.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2807, pruned_loss=0.05709, over 1606141.23 frames. ], batch size: 28, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:24:38,876 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211774.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:24:59,721 INFO [train.py:901] (1/4) Epoch 27, batch 1650, loss[loss=0.192, simple_loss=0.2864, pruned_loss=0.04874, over 8485.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2804, pruned_loss=0.05686, over 1610471.38 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:25:03,803 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.547e+02 3.089e+02 3.889e+02 1.356e+03, threshold=6.177e+02, percent-clipped=3.0 +2023-02-07 11:25:15,707 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4419, 1.8601, 2.8948, 1.3239, 2.1210, 1.8023, 1.5669, 2.1581], + device='cuda:1'), covar=tensor([0.2425, 0.3040, 0.1133, 0.5477, 0.2365, 0.4013, 0.2976, 0.2870], + device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0634, 0.0566, 0.0670, 0.0659, 0.0610, 0.0562, 0.0648], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:25:18,306 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.7661, 5.9538, 5.1944, 2.8262, 5.1873, 5.6884, 5.3298, 5.4350], + device='cuda:1'), covar=tensor([0.0516, 0.0404, 0.0916, 0.3902, 0.0780, 0.0631, 0.1095, 0.0456], + device='cuda:1'), in_proj_covar=tensor([0.0543, 0.0459, 0.0445, 0.0557, 0.0441, 0.0463, 0.0437, 0.0406], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:25:34,300 INFO [train.py:901] (1/4) Epoch 27, batch 1700, loss[loss=0.2014, simple_loss=0.289, pruned_loss=0.05692, over 8552.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2796, pruned_loss=0.05566, over 1613032.93 frames. ], batch size: 31, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:25:39,905 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211863.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:25:57,119 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9663, 1.6612, 3.4385, 1.4406, 2.2727, 3.8426, 3.9705, 3.3060], + device='cuda:1'), covar=tensor([0.1252, 0.1790, 0.0337, 0.2312, 0.1215, 0.0248, 0.0601, 0.0525], + device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0329, 0.0295, 0.0322, 0.0323, 0.0279, 0.0440, 0.0308], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 11:25:59,114 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211889.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:26:09,570 INFO [train.py:901] (1/4) Epoch 27, batch 1750, loss[loss=0.197, simple_loss=0.2783, pruned_loss=0.05783, over 7700.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2803, pruned_loss=0.05577, over 1613738.80 frames. ], batch size: 18, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:26:13,486 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.324e+02 2.460e+02 2.972e+02 3.773e+02 5.726e+02, threshold=5.944e+02, percent-clipped=0.0 +2023-02-07 11:26:24,664 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.81 vs. limit=5.0 +2023-02-07 11:26:43,427 INFO [train.py:901] (1/4) Epoch 27, batch 1800, loss[loss=0.1749, simple_loss=0.2598, pruned_loss=0.045, over 7824.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2794, pruned_loss=0.05551, over 1609290.15 frames. ], batch size: 20, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:27:03,805 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-02-07 11:27:20,494 INFO [train.py:901] (1/4) Epoch 27, batch 1850, loss[loss=0.159, simple_loss=0.2283, pruned_loss=0.04479, over 7704.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2793, pruned_loss=0.05557, over 1608875.87 frames. ], batch size: 18, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:27:24,580 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.252e+02 2.767e+02 3.484e+02 5.487e+02, threshold=5.534e+02, percent-clipped=0.0 +2023-02-07 11:27:54,169 INFO [train.py:901] (1/4) Epoch 27, batch 1900, loss[loss=0.1604, simple_loss=0.243, pruned_loss=0.03889, over 8286.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2792, pruned_loss=0.05576, over 1609571.49 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:28:23,230 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.6699, 5.7336, 5.0472, 2.7792, 5.0263, 5.4630, 5.1964, 5.2863], + device='cuda:1'), covar=tensor([0.0511, 0.0418, 0.0951, 0.3970, 0.0778, 0.0891, 0.1104, 0.0517], + device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0456, 0.0442, 0.0552, 0.0438, 0.0460, 0.0437, 0.0402], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:28:25,184 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 11:28:28,384 INFO [train.py:901] (1/4) Epoch 27, batch 1950, loss[loss=0.1833, simple_loss=0.2618, pruned_loss=0.05239, over 8228.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2795, pruned_loss=0.05562, over 1610920.55 frames. ], batch size: 22, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:28:33,122 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.484e+02 3.059e+02 3.727e+02 7.478e+02, threshold=6.119e+02, percent-clipped=3.0 +2023-02-07 11:28:39,154 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 11:28:39,365 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212119.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:28:41,242 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212122.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 11:28:56,768 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212144.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:28:57,280 WARNING [train.py:1067] (1/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 11:28:57,489 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212145.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:29:03,971 INFO [train.py:901] (1/4) Epoch 27, batch 2000, loss[loss=0.201, simple_loss=0.2944, pruned_loss=0.05379, over 8197.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2815, pruned_loss=0.05644, over 1619595.25 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:29:14,165 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212170.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:29:37,633 INFO [train.py:901] (1/4) Epoch 27, batch 2050, loss[loss=0.224, simple_loss=0.3063, pruned_loss=0.07089, over 8648.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2817, pruned_loss=0.05674, over 1621818.29 frames. ], batch size: 49, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:29:41,738 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.364e+02 2.966e+02 3.655e+02 9.314e+02, threshold=5.932e+02, percent-clipped=4.0 +2023-02-07 11:29:53,462 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5801, 1.3535, 1.5483, 1.3314, 0.9748, 1.3994, 1.3997, 1.2434], + device='cuda:1'), covar=tensor([0.0617, 0.1310, 0.1761, 0.1540, 0.0612, 0.1510, 0.0778, 0.0730], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0154, 0.0190, 0.0163, 0.0102, 0.0164, 0.0113, 0.0146], + device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 11:30:13,815 INFO [train.py:901] (1/4) Epoch 27, batch 2100, loss[loss=0.1629, simple_loss=0.2435, pruned_loss=0.04112, over 7235.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2816, pruned_loss=0.05686, over 1622268.70 frames. ], batch size: 16, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:30:24,389 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2786, 3.1172, 2.9793, 1.4808, 2.8868, 2.9338, 2.8660, 2.8572], + device='cuda:1'), covar=tensor([0.1067, 0.0825, 0.1258, 0.4345, 0.1011, 0.1300, 0.1544, 0.1092], + device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0456, 0.0442, 0.0554, 0.0438, 0.0459, 0.0437, 0.0404], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:30:25,126 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212271.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:30:42,129 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-02-07 11:30:47,851 INFO [train.py:901] (1/4) Epoch 27, batch 2150, loss[loss=0.1783, simple_loss=0.2584, pruned_loss=0.04916, over 8240.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2814, pruned_loss=0.05649, over 1620476.24 frames. ], batch size: 22, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:30:48,019 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0268, 1.6461, 1.4405, 1.5653, 1.3694, 1.3074, 1.3058, 1.2648], + device='cuda:1'), covar=tensor([0.1185, 0.0564, 0.1368, 0.0615, 0.0749, 0.1614, 0.0942, 0.0867], + device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0240, 0.0341, 0.0316, 0.0304, 0.0349, 0.0350, 0.0323], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 11:30:51,761 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.395e+02 2.764e+02 3.582e+02 6.444e+02, threshold=5.527e+02, percent-clipped=1.0 +2023-02-07 11:31:22,778 INFO [train.py:901] (1/4) Epoch 27, batch 2200, loss[loss=0.2251, simple_loss=0.3151, pruned_loss=0.06755, over 8472.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2823, pruned_loss=0.0574, over 1621667.94 frames. ], batch size: 25, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:31:57,347 INFO [train.py:901] (1/4) Epoch 27, batch 2250, loss[loss=0.2276, simple_loss=0.3195, pruned_loss=0.06787, over 8474.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2818, pruned_loss=0.05742, over 1618312.56 frames. ], batch size: 25, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:32:01,575 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.301e+02 2.815e+02 3.457e+02 5.141e+02, threshold=5.631e+02, percent-clipped=0.0 +2023-02-07 11:32:31,467 INFO [train.py:901] (1/4) Epoch 27, batch 2300, loss[loss=0.1831, simple_loss=0.2715, pruned_loss=0.04733, over 8477.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2823, pruned_loss=0.05736, over 1621954.00 frames. ], batch size: 29, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:32:32,833 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212457.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:32:39,376 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=212466.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 11:33:07,938 INFO [train.py:901] (1/4) Epoch 27, batch 2350, loss[loss=0.2559, simple_loss=0.3263, pruned_loss=0.09273, over 6920.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2824, pruned_loss=0.05734, over 1618716.99 frames. ], batch size: 71, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:33:12,148 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.444e+02 2.366e+02 2.801e+02 3.492e+02 6.818e+02, threshold=5.601e+02, percent-clipped=4.0 +2023-02-07 11:33:42,849 INFO [train.py:901] (1/4) Epoch 27, batch 2400, loss[loss=0.195, simple_loss=0.2898, pruned_loss=0.05007, over 8033.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2828, pruned_loss=0.0576, over 1620399.20 frames. ], batch size: 22, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:33:46,056 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 11:34:01,674 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212581.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 11:34:19,818 INFO [train.py:901] (1/4) Epoch 27, batch 2450, loss[loss=0.204, simple_loss=0.2891, pruned_loss=0.05948, over 8442.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2832, pruned_loss=0.05758, over 1623166.41 frames. ], batch size: 27, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:34:23,905 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.730e+02 2.406e+02 2.879e+02 3.948e+02 9.646e+02, threshold=5.757e+02, percent-clipped=9.0 +2023-02-07 11:34:26,762 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=212615.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:34:41,434 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 +2023-02-07 11:34:54,051 INFO [train.py:901] (1/4) Epoch 27, batch 2500, loss[loss=0.2842, simple_loss=0.3637, pruned_loss=0.1023, over 6940.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2833, pruned_loss=0.05838, over 1622105.44 frames. ], batch size: 71, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:35:15,631 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-07 11:35:21,446 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.4308, 2.1641, 2.6997, 2.2656, 2.8053, 2.4525, 2.3103, 1.5742], + device='cuda:1'), covar=tensor([0.5683, 0.4909, 0.2119, 0.4129, 0.2429, 0.3199, 0.1826, 0.5458], + device='cuda:1'), in_proj_covar=tensor([0.0966, 0.1018, 0.0830, 0.0986, 0.1023, 0.0926, 0.0770, 0.0847], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 11:35:28,282 INFO [train.py:901] (1/4) Epoch 27, batch 2550, loss[loss=0.2082, simple_loss=0.2993, pruned_loss=0.0586, over 8190.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2835, pruned_loss=0.05867, over 1620220.14 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:35:33,073 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.333e+02 2.985e+02 3.926e+02 7.498e+02, threshold=5.971e+02, percent-clipped=4.0 +2023-02-07 11:35:37,331 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212716.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:35:46,856 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212729.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:35:47,588 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212730.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:36:01,350 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5674, 1.8585, 1.9785, 1.2810, 2.0274, 1.4221, 0.6234, 1.7306], + device='cuda:1'), covar=tensor([0.0743, 0.0421, 0.0331, 0.0731, 0.0434, 0.1058, 0.1045, 0.0426], + device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0410, 0.0364, 0.0461, 0.0394, 0.0552, 0.0404, 0.0442], + device='cuda:1'), out_proj_covar=tensor([1.2576e-04, 1.0662e-04, 9.5043e-05, 1.2078e-04, 1.0322e-04, 1.5429e-04, + 1.0796e-04, 1.1593e-04], device='cuda:1') +2023-02-07 11:36:04,558 INFO [train.py:901] (1/4) Epoch 27, batch 2600, loss[loss=0.1815, simple_loss=0.2709, pruned_loss=0.04608, over 8433.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.283, pruned_loss=0.05783, over 1623638.19 frames. ], batch size: 27, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:36:05,343 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7288, 1.6459, 2.1096, 1.4470, 1.2801, 2.1315, 0.3397, 1.3712], + device='cuda:1'), covar=tensor([0.1459, 0.1065, 0.0375, 0.1022, 0.2347, 0.0366, 0.1789, 0.1206], + device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0205, 0.0134, 0.0223, 0.0277, 0.0144, 0.0172, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 11:36:29,683 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212792.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:36:35,820 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=212801.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:36:38,499 INFO [train.py:901] (1/4) Epoch 27, batch 2650, loss[loss=0.1587, simple_loss=0.2496, pruned_loss=0.03387, over 7795.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2836, pruned_loss=0.05846, over 1621329.08 frames. ], batch size: 19, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:36:43,310 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.517e+02 2.957e+02 3.589e+02 7.428e+02, threshold=5.913e+02, percent-clipped=3.0 +2023-02-07 11:37:02,117 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212837.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 11:37:03,989 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5675, 4.5476, 4.1280, 2.3167, 4.0594, 4.2870, 4.0567, 4.0476], + device='cuda:1'), covar=tensor([0.0712, 0.0486, 0.0965, 0.4196, 0.0866, 0.0767, 0.1243, 0.0638], + device='cuda:1'), in_proj_covar=tensor([0.0542, 0.0458, 0.0444, 0.0555, 0.0439, 0.0463, 0.0437, 0.0404], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:37:14,963 INFO [train.py:901] (1/4) Epoch 27, batch 2700, loss[loss=0.2123, simple_loss=0.3145, pruned_loss=0.05511, over 8747.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.284, pruned_loss=0.0586, over 1618868.15 frames. ], batch size: 30, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:37:19,862 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212862.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 11:37:49,376 INFO [train.py:901] (1/4) Epoch 27, batch 2750, loss[loss=0.1666, simple_loss=0.2562, pruned_loss=0.03849, over 7544.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2832, pruned_loss=0.05848, over 1614148.69 frames. ], batch size: 18, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:37:53,340 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.387e+02 2.942e+02 3.576e+02 8.277e+02, threshold=5.883e+02, percent-clipped=4.0 +2023-02-07 11:37:56,789 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212916.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:38:09,049 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5397, 2.1348, 3.1327, 1.4023, 2.3264, 1.9202, 1.6774, 2.3469], + device='cuda:1'), covar=tensor([0.1994, 0.2532, 0.0888, 0.4809, 0.2166, 0.3458, 0.2608, 0.2576], + device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0636, 0.0569, 0.0672, 0.0661, 0.0612, 0.0563, 0.0649], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:38:25,581 INFO [train.py:901] (1/4) Epoch 27, batch 2800, loss[loss=0.183, simple_loss=0.2823, pruned_loss=0.04185, over 8105.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2833, pruned_loss=0.05858, over 1615463.52 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:38:29,013 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6342, 1.8758, 1.9287, 1.8559, 1.1481, 1.7439, 2.2876, 1.9153], + device='cuda:1'), covar=tensor([0.0494, 0.1122, 0.1574, 0.1303, 0.0666, 0.1374, 0.0696, 0.0617], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0154, 0.0190, 0.0163, 0.0102, 0.0164, 0.0113, 0.0147], + device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 11:38:42,835 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9190, 2.0321, 1.7129, 2.5899, 1.1598, 1.5323, 1.9240, 2.0644], + device='cuda:1'), covar=tensor([0.0679, 0.0746, 0.0843, 0.0333, 0.1097, 0.1234, 0.0740, 0.0732], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0195, 0.0244, 0.0213, 0.0203, 0.0245, 0.0249, 0.0204], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 11:38:46,069 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212986.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:38:58,529 INFO [train.py:901] (1/4) Epoch 27, batch 2850, loss[loss=0.2254, simple_loss=0.305, pruned_loss=0.07293, over 7197.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2832, pruned_loss=0.05895, over 1611766.26 frames. ], batch size: 16, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:39:02,633 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.396e+02 2.420e+02 3.040e+02 3.738e+02 9.771e+02, threshold=6.080e+02, percent-clipped=4.0 +2023-02-07 11:39:02,868 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213011.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:39:19,518 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.2680, 2.5566, 2.7920, 1.6409, 3.0787, 1.8608, 1.4552, 2.1823], + device='cuda:1'), covar=tensor([0.0914, 0.0465, 0.0307, 0.0872, 0.0517, 0.0921, 0.1161, 0.0651], + device='cuda:1'), in_proj_covar=tensor([0.0471, 0.0408, 0.0361, 0.0459, 0.0394, 0.0548, 0.0402, 0.0439], + device='cuda:1'), out_proj_covar=tensor([1.2487e-04, 1.0615e-04, 9.4338e-05, 1.2007e-04, 1.0310e-04, 1.5314e-04, + 1.0733e-04, 1.1538e-04], device='cuda:1') +2023-02-07 11:39:33,423 INFO [train.py:901] (1/4) Epoch 27, batch 2900, loss[loss=0.1846, simple_loss=0.279, pruned_loss=0.04506, over 8335.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2823, pruned_loss=0.05847, over 1614315.75 frames. ], batch size: 26, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:39:33,647 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4326, 1.8671, 3.2682, 1.3096, 2.4671, 1.9666, 1.4762, 2.4312], + device='cuda:1'), covar=tensor([0.2323, 0.2827, 0.0829, 0.5218, 0.2049, 0.3318, 0.2800, 0.2421], + device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0635, 0.0568, 0.0670, 0.0659, 0.0610, 0.0562, 0.0648], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:39:36,853 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=213060.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:39:37,639 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.2403, 2.1532, 1.7251, 1.9383, 1.7402, 1.4622, 1.7123, 1.6038], + device='cuda:1'), covar=tensor([0.1296, 0.0414, 0.1175, 0.0545, 0.0750, 0.1505, 0.0913, 0.0913], + device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0241, 0.0340, 0.0315, 0.0304, 0.0347, 0.0349, 0.0323], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 11:39:46,825 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=213073.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:40:09,155 INFO [train.py:901] (1/4) Epoch 27, batch 2950, loss[loss=0.2533, simple_loss=0.3197, pruned_loss=0.09347, over 7214.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2835, pruned_loss=0.05906, over 1617256.83 frames. ], batch size: 71, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:40:12,519 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 11:40:13,187 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.292e+02 2.734e+02 3.601e+02 6.803e+02, threshold=5.467e+02, percent-clipped=1.0 +2023-02-07 11:40:30,259 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=213136.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:40:42,822 INFO [train.py:901] (1/4) Epoch 27, batch 3000, loss[loss=0.1727, simple_loss=0.2424, pruned_loss=0.05152, over 7927.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.283, pruned_loss=0.05834, over 1618484.02 frames. ], batch size: 20, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:40:42,822 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 11:40:51,047 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7791, 1.7434, 1.5844, 2.2451, 1.1178, 1.4881, 1.7885, 1.7988], + device='cuda:1'), covar=tensor([0.0733, 0.0906, 0.0923, 0.0447, 0.1154, 0.1364, 0.0736, 0.0817], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0195, 0.0243, 0.0212, 0.0203, 0.0245, 0.0249, 0.0203], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 11:40:56,478 INFO [train.py:935] (1/4) Epoch 27, validation: loss=0.171, simple_loss=0.2706, pruned_loss=0.03572, over 944034.00 frames. +2023-02-07 11:40:56,478 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 11:41:08,336 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213172.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:41:10,344 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213175.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:41:19,864 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213188.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:41:20,565 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.5407, 1.3457, 4.7335, 1.6072, 4.1940, 3.8714, 4.2727, 4.1419], + device='cuda:1'), covar=tensor([0.0618, 0.5045, 0.0463, 0.4653, 0.1054, 0.1022, 0.0595, 0.0679], + device='cuda:1'), in_proj_covar=tensor([0.0676, 0.0668, 0.0736, 0.0660, 0.0749, 0.0638, 0.0638, 0.0719], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:41:25,953 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213197.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:41:31,706 INFO [train.py:901] (1/4) Epoch 27, batch 3050, loss[loss=0.1916, simple_loss=0.2872, pruned_loss=0.04793, over 8517.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2843, pruned_loss=0.0588, over 1619844.97 frames. ], batch size: 28, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:41:36,535 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.283e+02 2.877e+02 3.649e+02 6.604e+02, threshold=5.754e+02, percent-clipped=7.0 +2023-02-07 11:41:46,123 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7592, 1.5963, 2.2264, 1.3711, 1.3033, 2.2246, 0.5635, 1.4126], + device='cuda:1'), covar=tensor([0.1411, 0.1126, 0.0370, 0.1166, 0.2276, 0.0373, 0.1825, 0.1320], + device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0203, 0.0134, 0.0223, 0.0275, 0.0144, 0.0171, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 11:42:01,589 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 11:42:03,697 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 +2023-02-07 11:42:04,102 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213251.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:42:06,027 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9583, 1.4209, 1.7152, 1.3470, 0.9868, 1.4642, 1.8522, 1.4540], + device='cuda:1'), covar=tensor([0.0546, 0.1299, 0.1708, 0.1489, 0.0602, 0.1587, 0.0674, 0.0723], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0190, 0.0162, 0.0101, 0.0164, 0.0112, 0.0146], + device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 11:42:06,437 INFO [train.py:901] (1/4) Epoch 27, batch 3100, loss[loss=0.2504, simple_loss=0.3184, pruned_loss=0.0912, over 7802.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2838, pruned_loss=0.05877, over 1618672.44 frames. ], batch size: 20, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:42:07,838 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1004, 1.2217, 4.3239, 2.0548, 2.6308, 4.8972, 4.9791, 4.2220], + device='cuda:1'), covar=tensor([0.1227, 0.2136, 0.0293, 0.1917, 0.1114, 0.0171, 0.0380, 0.0550], + device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0327, 0.0293, 0.0321, 0.0321, 0.0279, 0.0438, 0.0309], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 11:42:18,672 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-07 11:42:39,679 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9358, 2.0629, 1.6857, 2.7513, 1.1882, 1.5437, 2.0368, 2.1150], + device='cuda:1'), covar=tensor([0.0745, 0.0789, 0.0929, 0.0332, 0.1225, 0.1386, 0.0800, 0.0828], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0197, 0.0246, 0.0214, 0.0205, 0.0248, 0.0252, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 11:42:40,174 INFO [train.py:901] (1/4) Epoch 27, batch 3150, loss[loss=0.1802, simple_loss=0.2763, pruned_loss=0.04202, over 8470.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2846, pruned_loss=0.0589, over 1619355.91 frames. ], batch size: 25, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:42:44,227 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.764e+02 2.557e+02 3.186e+02 3.836e+02 1.080e+03, threshold=6.372e+02, percent-clipped=6.0 +2023-02-07 11:43:15,308 INFO [train.py:901] (1/4) Epoch 27, batch 3200, loss[loss=0.2074, simple_loss=0.2804, pruned_loss=0.06719, over 7971.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2862, pruned_loss=0.05937, over 1625255.53 frames. ], batch size: 21, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:43:48,887 INFO [train.py:901] (1/4) Epoch 27, batch 3250, loss[loss=0.2096, simple_loss=0.2975, pruned_loss=0.06085, over 8497.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2841, pruned_loss=0.05841, over 1619579.30 frames. ], batch size: 29, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:43:52,812 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.478e+02 2.885e+02 3.413e+02 5.983e+02, threshold=5.770e+02, percent-clipped=0.0 +2023-02-07 11:44:07,276 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213431.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:44:09,272 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8699, 1.4487, 3.1416, 1.4792, 2.4133, 3.3899, 3.5060, 2.8881], + device='cuda:1'), covar=tensor([0.1241, 0.1908, 0.0335, 0.2244, 0.0928, 0.0259, 0.0691, 0.0556], + device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0328, 0.0294, 0.0322, 0.0323, 0.0280, 0.0440, 0.0310], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 11:44:17,674 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213444.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:44:25,458 INFO [train.py:901] (1/4) Epoch 27, batch 3300, loss[loss=0.2327, simple_loss=0.3073, pruned_loss=0.07907, over 8460.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2847, pruned_loss=0.05884, over 1615384.67 frames. ], batch size: 27, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:44:26,271 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213456.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:44:35,226 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213469.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:44:52,790 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0191, 1.7529, 2.5047, 1.7745, 1.4616, 2.5336, 0.5979, 1.5853], + device='cuda:1'), covar=tensor([0.1286, 0.1232, 0.0309, 0.1018, 0.2241, 0.0311, 0.1726, 0.1217], + device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0203, 0.0134, 0.0223, 0.0275, 0.0144, 0.0171, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 11:44:59,357 INFO [train.py:901] (1/4) Epoch 27, batch 3350, loss[loss=0.2422, simple_loss=0.3326, pruned_loss=0.07592, over 8192.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2848, pruned_loss=0.0588, over 1618712.11 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:45:00,954 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213507.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:45:03,449 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.606e+02 3.102e+02 3.998e+02 8.787e+02, threshold=6.203e+02, percent-clipped=8.0 +2023-02-07 11:45:18,451 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213532.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:45:34,517 INFO [train.py:901] (1/4) Epoch 27, batch 3400, loss[loss=0.1684, simple_loss=0.2493, pruned_loss=0.04372, over 7688.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.284, pruned_loss=0.05835, over 1616506.75 frames. ], batch size: 18, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:45:55,323 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213584.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:46:09,626 INFO [train.py:901] (1/4) Epoch 27, batch 3450, loss[loss=0.1929, simple_loss=0.2816, pruned_loss=0.05212, over 8328.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.283, pruned_loss=0.05769, over 1616575.90 frames. ], batch size: 25, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:46:13,703 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.444e+02 2.297e+02 2.616e+02 3.439e+02 9.820e+02, threshold=5.232e+02, percent-clipped=1.0 +2023-02-07 11:46:44,526 INFO [train.py:901] (1/4) Epoch 27, batch 3500, loss[loss=0.1793, simple_loss=0.2533, pruned_loss=0.05262, over 7440.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2827, pruned_loss=0.05791, over 1614698.03 frames. ], batch size: 17, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:46:46,128 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4989, 1.6481, 2.1723, 1.4074, 1.5222, 1.7373, 1.5339, 1.5415], + device='cuda:1'), covar=tensor([0.2078, 0.2684, 0.1041, 0.4790, 0.2078, 0.3671, 0.2612, 0.2257], + device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0634, 0.0568, 0.0671, 0.0660, 0.0609, 0.0562, 0.0647], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:47:08,392 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.5560, 5.5550, 4.9708, 2.7779, 4.9241, 5.2975, 5.0338, 5.0370], + device='cuda:1'), covar=tensor([0.0504, 0.0386, 0.0768, 0.3543, 0.0731, 0.0894, 0.1079, 0.0593], + device='cuda:1'), in_proj_covar=tensor([0.0541, 0.0459, 0.0446, 0.0558, 0.0441, 0.0465, 0.0439, 0.0405], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:47:08,657 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 +2023-02-07 11:47:11,033 WARNING [train.py:1067] (1/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 11:47:20,454 INFO [train.py:901] (1/4) Epoch 27, batch 3550, loss[loss=0.2369, simple_loss=0.3173, pruned_loss=0.07825, over 8191.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2832, pruned_loss=0.05844, over 1610340.49 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:47:24,349 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 2.484e+02 3.157e+02 3.893e+02 8.912e+02, threshold=6.313e+02, percent-clipped=7.0 +2023-02-07 11:47:51,925 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7781, 1.5868, 2.3986, 1.5625, 1.2961, 2.3880, 0.4985, 1.4651], + device='cuda:1'), covar=tensor([0.1467, 0.1359, 0.0319, 0.1083, 0.2643, 0.0347, 0.1944, 0.1312], + device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0206, 0.0135, 0.0224, 0.0277, 0.0145, 0.0172, 0.0200], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 11:47:55,110 INFO [train.py:901] (1/4) Epoch 27, batch 3600, loss[loss=0.1515, simple_loss=0.2339, pruned_loss=0.03452, over 7722.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2829, pruned_loss=0.05823, over 1612949.98 frames. ], batch size: 18, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:48:31,566 INFO [train.py:901] (1/4) Epoch 27, batch 3650, loss[loss=0.1782, simple_loss=0.2542, pruned_loss=0.05113, over 7541.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2821, pruned_loss=0.05776, over 1612444.77 frames. ], batch size: 18, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:48:35,659 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.435e+02 3.005e+02 4.000e+02 1.001e+03, threshold=6.009e+02, percent-clipped=1.0 +2023-02-07 11:49:05,215 INFO [train.py:901] (1/4) Epoch 27, batch 3700, loss[loss=0.225, simple_loss=0.3042, pruned_loss=0.07292, over 8596.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2822, pruned_loss=0.05757, over 1612281.10 frames. ], batch size: 31, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:49:11,355 WARNING [train.py:1067] (1/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 11:49:27,281 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 11:49:40,754 INFO [train.py:901] (1/4) Epoch 27, batch 3750, loss[loss=0.189, simple_loss=0.2752, pruned_loss=0.05139, over 8241.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2833, pruned_loss=0.05789, over 1613647.95 frames. ], batch size: 22, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:49:44,686 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.240e+02 2.670e+02 3.453e+02 6.024e+02, threshold=5.340e+02, percent-clipped=1.0 +2023-02-07 11:49:57,382 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=213928.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:50:07,450 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.3346, 2.0771, 2.1296, 1.9900, 1.4672, 2.0628, 2.2577, 2.0445], + device='cuda:1'), covar=tensor([0.0596, 0.0956, 0.1356, 0.1188, 0.0639, 0.1170, 0.0748, 0.0539], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0154, 0.0190, 0.0162, 0.0101, 0.0164, 0.0113, 0.0146], + device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 11:50:15,109 INFO [train.py:901] (1/4) Epoch 27, batch 3800, loss[loss=0.1969, simple_loss=0.2901, pruned_loss=0.05182, over 8341.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2822, pruned_loss=0.05783, over 1609507.90 frames. ], batch size: 26, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:50:18,681 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6579, 2.1116, 3.1535, 1.5111, 2.3168, 2.1124, 1.7972, 2.4191], + device='cuda:1'), covar=tensor([0.2005, 0.2546, 0.0926, 0.4679, 0.2031, 0.3200, 0.2418, 0.2288], + device='cuda:1'), in_proj_covar=tensor([0.0537, 0.0632, 0.0565, 0.0667, 0.0659, 0.0606, 0.0561, 0.0644], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:50:51,703 INFO [train.py:901] (1/4) Epoch 27, batch 3850, loss[loss=0.1592, simple_loss=0.2521, pruned_loss=0.03316, over 7527.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.282, pruned_loss=0.05746, over 1609367.52 frames. ], batch size: 18, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:50:55,678 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 2.325e+02 2.987e+02 3.815e+02 9.366e+02, threshold=5.974e+02, percent-clipped=6.0 +2023-02-07 11:51:14,749 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.6992, 2.5326, 3.3120, 2.6810, 3.2848, 2.7309, 2.6035, 2.0374], + device='cuda:1'), covar=tensor([0.5559, 0.5353, 0.1956, 0.4027, 0.2537, 0.3100, 0.1765, 0.5808], + device='cuda:1'), in_proj_covar=tensor([0.0961, 0.1014, 0.0828, 0.0985, 0.1017, 0.0923, 0.0765, 0.0845], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 11:51:19,438 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214043.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:51:21,269 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 11:51:27,208 INFO [train.py:901] (1/4) Epoch 27, batch 3900, loss[loss=0.1964, simple_loss=0.2893, pruned_loss=0.05172, over 8481.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.282, pruned_loss=0.05746, over 1613639.42 frames. ], batch size: 27, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:51:39,612 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-07 11:52:00,518 INFO [train.py:901] (1/4) Epoch 27, batch 3950, loss[loss=0.1759, simple_loss=0.252, pruned_loss=0.04988, over 7442.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2821, pruned_loss=0.05734, over 1615929.30 frames. ], batch size: 17, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:52:04,373 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.377e+02 2.717e+02 3.364e+02 5.097e+02, threshold=5.435e+02, percent-clipped=0.0 +2023-02-07 11:52:13,417 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-07 11:52:13,847 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.8594, 2.0900, 3.5487, 1.8399, 1.9450, 3.5608, 0.7296, 2.0346], + device='cuda:1'), covar=tensor([0.1145, 0.1202, 0.0238, 0.1519, 0.2140, 0.0258, 0.2029, 0.1327], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0206, 0.0136, 0.0225, 0.0279, 0.0146, 0.0173, 0.0200], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 11:52:30,743 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214147.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:52:35,674 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.27 vs. limit=5.0 +2023-02-07 11:52:36,018 INFO [train.py:901] (1/4) Epoch 27, batch 4000, loss[loss=0.1714, simple_loss=0.2687, pruned_loss=0.03709, over 8026.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.282, pruned_loss=0.05716, over 1617239.31 frames. ], batch size: 22, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:53:10,789 INFO [train.py:901] (1/4) Epoch 27, batch 4050, loss[loss=0.2019, simple_loss=0.2808, pruned_loss=0.06152, over 7789.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2813, pruned_loss=0.05721, over 1613924.76 frames. ], batch size: 19, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:53:14,931 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.649e+02 2.379e+02 2.958e+02 3.648e+02 7.596e+02, threshold=5.915e+02, percent-clipped=3.0 +2023-02-07 11:53:31,871 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214236.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:53:45,409 INFO [train.py:901] (1/4) Epoch 27, batch 4100, loss[loss=0.2269, simple_loss=0.3111, pruned_loss=0.07137, over 8600.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2828, pruned_loss=0.05776, over 1616750.85 frames. ], batch size: 31, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:54:17,263 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214299.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:54:21,106 INFO [train.py:901] (1/4) Epoch 27, batch 4150, loss[loss=0.1648, simple_loss=0.2492, pruned_loss=0.04019, over 8085.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2825, pruned_loss=0.05778, over 1613814.61 frames. ], batch size: 21, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:54:25,184 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.522e+02 2.957e+02 3.518e+02 6.524e+02, threshold=5.913e+02, percent-clipped=2.0 +2023-02-07 11:54:34,220 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214324.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:54:55,406 INFO [train.py:901] (1/4) Epoch 27, batch 4200, loss[loss=0.1937, simple_loss=0.2701, pruned_loss=0.05863, over 7919.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2817, pruned_loss=0.05719, over 1608301.35 frames. ], batch size: 20, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:55:15,711 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 11:55:30,770 INFO [train.py:901] (1/4) Epoch 27, batch 4250, loss[loss=0.1829, simple_loss=0.2709, pruned_loss=0.0474, over 8241.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2825, pruned_loss=0.05766, over 1611210.27 frames. ], batch size: 22, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:55:34,791 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.351e+02 2.930e+02 3.605e+02 8.966e+02, threshold=5.860e+02, percent-clipped=4.0 +2023-02-07 11:55:40,136 WARNING [train.py:1067] (1/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 11:56:04,680 INFO [train.py:901] (1/4) Epoch 27, batch 4300, loss[loss=0.1977, simple_loss=0.2836, pruned_loss=0.05594, over 8501.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2827, pruned_loss=0.0585, over 1611451.56 frames. ], batch size: 26, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:56:19,497 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6348, 2.1594, 3.2659, 1.5293, 2.4691, 2.0728, 1.7466, 2.4281], + device='cuda:1'), covar=tensor([0.2003, 0.2643, 0.0908, 0.4793, 0.2012, 0.3399, 0.2593, 0.2467], + device='cuda:1'), in_proj_covar=tensor([0.0541, 0.0635, 0.0566, 0.0671, 0.0662, 0.0611, 0.0564, 0.0647], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:56:29,281 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=214491.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:56:40,083 INFO [train.py:901] (1/4) Epoch 27, batch 4350, loss[loss=0.1678, simple_loss=0.2441, pruned_loss=0.04575, over 7701.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2837, pruned_loss=0.05861, over 1615063.57 frames. ], batch size: 18, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:56:44,887 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.529e+02 2.457e+02 2.989e+02 4.041e+02 8.697e+02, threshold=5.978e+02, percent-clipped=4.0 +2023-02-07 11:57:11,488 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 11:57:14,874 INFO [train.py:901] (1/4) Epoch 27, batch 4400, loss[loss=0.2133, simple_loss=0.3035, pruned_loss=0.06159, over 7969.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2828, pruned_loss=0.05821, over 1611433.45 frames. ], batch size: 21, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:57:25,505 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6746, 2.1404, 3.3154, 1.6259, 2.4203, 2.1600, 1.8026, 2.4893], + device='cuda:1'), covar=tensor([0.2042, 0.2755, 0.0920, 0.4698, 0.2043, 0.3280, 0.2596, 0.2514], + device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0632, 0.0563, 0.0668, 0.0658, 0.0607, 0.0562, 0.0643], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 11:57:32,006 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=214580.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:57:39,612 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 11:57:43,110 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 +2023-02-07 11:57:47,096 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8269, 2.0516, 2.1513, 1.5759, 2.3482, 1.5861, 0.7571, 2.1532], + device='cuda:1'), covar=tensor([0.0742, 0.0434, 0.0345, 0.0633, 0.0414, 0.0988, 0.1067, 0.0301], + device='cuda:1'), in_proj_covar=tensor([0.0472, 0.0409, 0.0362, 0.0459, 0.0394, 0.0549, 0.0402, 0.0438], + device='cuda:1'), out_proj_covar=tensor([1.2508e-04, 1.0637e-04, 9.4530e-05, 1.2029e-04, 1.0328e-04, 1.5321e-04, + 1.0741e-04, 1.1485e-04], device='cuda:1') +2023-02-07 11:57:49,419 INFO [train.py:901] (1/4) Epoch 27, batch 4450, loss[loss=0.2631, simple_loss=0.3471, pruned_loss=0.08952, over 8528.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2831, pruned_loss=0.0582, over 1618145.40 frames. ], batch size: 49, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:57:50,290 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214606.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:57:51,445 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 11:57:53,331 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.435e+02 2.910e+02 3.675e+02 1.096e+03, threshold=5.821e+02, percent-clipped=3.0 +2023-02-07 11:58:25,066 INFO [train.py:901] (1/4) Epoch 27, batch 4500, loss[loss=0.2078, simple_loss=0.298, pruned_loss=0.05877, over 8472.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2829, pruned_loss=0.05793, over 1619304.37 frames. ], batch size: 25, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:58:49,089 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 11:58:51,965 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214695.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:58:58,563 INFO [train.py:901] (1/4) Epoch 27, batch 4550, loss[loss=0.2038, simple_loss=0.2928, pruned_loss=0.05741, over 8290.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2824, pruned_loss=0.05757, over 1617175.68 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:59:03,195 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.415e+02 2.259e+02 2.795e+02 3.667e+02 7.490e+02, threshold=5.591e+02, percent-clipped=6.0 +2023-02-07 11:59:34,575 INFO [train.py:901] (1/4) Epoch 27, batch 4600, loss[loss=0.191, simple_loss=0.2727, pruned_loss=0.05462, over 8184.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2822, pruned_loss=0.05786, over 1610825.66 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 12:00:07,865 INFO [train.py:901] (1/4) Epoch 27, batch 4650, loss[loss=0.1952, simple_loss=0.2823, pruned_loss=0.05402, over 8569.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2826, pruned_loss=0.05794, over 1609592.98 frames. ], batch size: 31, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 12:00:11,922 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.452e+02 3.083e+02 3.974e+02 1.018e+03, threshold=6.165e+02, percent-clipped=5.0 +2023-02-07 12:00:43,658 INFO [train.py:901] (1/4) Epoch 27, batch 4700, loss[loss=0.1991, simple_loss=0.2923, pruned_loss=0.05294, over 8186.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.283, pruned_loss=0.05862, over 1607604.03 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 12:00:48,572 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214862.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:01:05,464 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214887.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:01:17,180 INFO [train.py:901] (1/4) Epoch 27, batch 4750, loss[loss=0.2331, simple_loss=0.3102, pruned_loss=0.078, over 8253.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2841, pruned_loss=0.05907, over 1610012.73 frames. ], batch size: 24, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 12:01:21,121 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.799e+02 2.444e+02 3.016e+02 3.790e+02 1.117e+03, threshold=6.032e+02, percent-clipped=6.0 +2023-02-07 12:01:42,269 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 12:01:45,086 WARNING [train.py:1067] (1/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 12:01:49,024 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214951.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:01:51,313 INFO [train.py:901] (1/4) Epoch 27, batch 4800, loss[loss=0.1735, simple_loss=0.2567, pruned_loss=0.04518, over 7640.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2826, pruned_loss=0.05866, over 1611297.04 frames. ], batch size: 19, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:02:07,893 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214976.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:02:27,034 INFO [train.py:901] (1/4) Epoch 27, batch 4850, loss[loss=0.2094, simple_loss=0.2832, pruned_loss=0.06784, over 7806.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.283, pruned_loss=0.05817, over 1617808.69 frames. ], batch size: 20, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:02:31,201 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.294e+02 2.705e+02 3.274e+02 6.085e+02, threshold=5.409e+02, percent-clipped=1.0 +2023-02-07 12:02:36,670 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 12:03:00,130 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5758, 2.1446, 3.1935, 1.5119, 2.4888, 2.0320, 1.6865, 2.5303], + device='cuda:1'), covar=tensor([0.2082, 0.2614, 0.0902, 0.4898, 0.1889, 0.3391, 0.2639, 0.2287], + device='cuda:1'), in_proj_covar=tensor([0.0541, 0.0636, 0.0567, 0.0671, 0.0663, 0.0613, 0.0566, 0.0648], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 12:03:01,830 INFO [train.py:901] (1/4) Epoch 27, batch 4900, loss[loss=0.254, simple_loss=0.3156, pruned_loss=0.09619, over 8464.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2829, pruned_loss=0.05745, over 1621088.61 frames. ], batch size: 25, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:03:37,419 INFO [train.py:901] (1/4) Epoch 27, batch 4950, loss[loss=0.2101, simple_loss=0.2949, pruned_loss=0.06262, over 8503.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2824, pruned_loss=0.05699, over 1613886.56 frames. ], batch size: 26, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:03:41,324 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.323e+02 2.858e+02 3.502e+02 9.819e+02, threshold=5.716e+02, percent-clipped=5.0 +2023-02-07 12:04:10,560 INFO [train.py:901] (1/4) Epoch 27, batch 5000, loss[loss=0.2049, simple_loss=0.2932, pruned_loss=0.05834, over 8476.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2838, pruned_loss=0.0577, over 1613998.87 frames. ], batch size: 25, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:04:47,122 INFO [train.py:901] (1/4) Epoch 27, batch 5050, loss[loss=0.2183, simple_loss=0.3017, pruned_loss=0.06745, over 8365.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2833, pruned_loss=0.05753, over 1615512.07 frames. ], batch size: 24, lr: 2.78e-03, grad_scale: 32.0 +2023-02-07 12:04:50,998 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.415e+02 2.920e+02 3.667e+02 5.760e+02, threshold=5.840e+02, percent-clipped=1.0 +2023-02-07 12:05:10,163 WARNING [train.py:1067] (1/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 12:05:15,552 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215248.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:05:20,048 INFO [train.py:901] (1/4) Epoch 27, batch 5100, loss[loss=0.2159, simple_loss=0.2896, pruned_loss=0.07114, over 8249.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2846, pruned_loss=0.05856, over 1616799.95 frames. ], batch size: 24, lr: 2.78e-03, grad_scale: 32.0 +2023-02-07 12:05:28,223 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8746, 1.6814, 2.3598, 1.5022, 1.4392, 2.3232, 0.8116, 1.6516], + device='cuda:1'), covar=tensor([0.1429, 0.1099, 0.0295, 0.1025, 0.2144, 0.0330, 0.1766, 0.1211], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0206, 0.0137, 0.0225, 0.0279, 0.0146, 0.0174, 0.0201], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 12:05:32,898 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5182, 1.3853, 1.8277, 1.3052, 1.1464, 1.7895, 0.2923, 1.2305], + device='cuda:1'), covar=tensor([0.1540, 0.1205, 0.0380, 0.0746, 0.2438, 0.0441, 0.1941, 0.1237], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0207, 0.0137, 0.0225, 0.0279, 0.0147, 0.0174, 0.0201], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 12:05:54,329 INFO [train.py:901] (1/4) Epoch 27, batch 5150, loss[loss=0.1827, simple_loss=0.2666, pruned_loss=0.04934, over 8298.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2824, pruned_loss=0.05728, over 1617410.51 frames. ], batch size: 23, lr: 2.78e-03, grad_scale: 32.0 +2023-02-07 12:05:59,267 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 2.449e+02 2.868e+02 3.492e+02 6.640e+02, threshold=5.736e+02, percent-clipped=1.0 +2023-02-07 12:06:22,713 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215343.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:06:28,157 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215351.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:06:30,696 INFO [train.py:901] (1/4) Epoch 27, batch 5200, loss[loss=0.2275, simple_loss=0.3164, pruned_loss=0.06928, over 8493.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.283, pruned_loss=0.05738, over 1620497.53 frames. ], batch size: 28, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:07:05,198 INFO [train.py:901] (1/4) Epoch 27, batch 5250, loss[loss=0.1784, simple_loss=0.2617, pruned_loss=0.04754, over 7929.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2829, pruned_loss=0.05702, over 1617326.69 frames. ], batch size: 20, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:07:09,825 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.358e+02 2.790e+02 3.638e+02 8.125e+02, threshold=5.579e+02, percent-clipped=3.0 +2023-02-07 12:07:12,624 WARNING [train.py:1067] (1/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 12:07:40,339 INFO [train.py:901] (1/4) Epoch 27, batch 5300, loss[loss=0.2123, simple_loss=0.2963, pruned_loss=0.0641, over 8232.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2828, pruned_loss=0.05705, over 1617617.25 frames. ], batch size: 22, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:08:13,771 INFO [train.py:901] (1/4) Epoch 27, batch 5350, loss[loss=0.1989, simple_loss=0.2844, pruned_loss=0.05664, over 8587.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2823, pruned_loss=0.05735, over 1609252.13 frames. ], batch size: 49, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:08:18,670 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 2.455e+02 2.847e+02 3.988e+02 1.267e+03, threshold=5.693e+02, percent-clipped=12.0 +2023-02-07 12:08:25,729 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215521.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:08:48,875 INFO [train.py:901] (1/4) Epoch 27, batch 5400, loss[loss=0.2557, simple_loss=0.3278, pruned_loss=0.09183, over 8499.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2824, pruned_loss=0.05741, over 1612686.31 frames. ], batch size: 29, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:08:52,411 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4966, 2.4910, 1.8236, 2.2189, 2.2094, 1.4453, 2.0527, 2.0407], + device='cuda:1'), covar=tensor([0.1573, 0.0400, 0.1201, 0.0656, 0.0732, 0.1667, 0.1063, 0.1069], + device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0241, 0.0343, 0.0312, 0.0303, 0.0347, 0.0350, 0.0323], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 12:08:57,736 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215566.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:09:08,905 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215583.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:09:14,846 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215592.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:09:23,341 INFO [train.py:901] (1/4) Epoch 27, batch 5450, loss[loss=0.1563, simple_loss=0.2411, pruned_loss=0.03579, over 7922.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2833, pruned_loss=0.05783, over 1618839.75 frames. ], batch size: 20, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:09:27,901 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 2.540e+02 3.136e+02 3.819e+02 8.555e+02, threshold=6.272e+02, percent-clipped=5.0 +2023-02-07 12:09:47,129 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8650, 1.5088, 1.7319, 1.3746, 0.8915, 1.5518, 1.6271, 1.6144], + device='cuda:1'), covar=tensor([0.0572, 0.1213, 0.1577, 0.1441, 0.0603, 0.1418, 0.0695, 0.0613], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0189, 0.0161, 0.0101, 0.0162, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 12:09:56,879 INFO [train.py:901] (1/4) Epoch 27, batch 5500, loss[loss=0.2292, simple_loss=0.3018, pruned_loss=0.07833, over 7101.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2843, pruned_loss=0.05868, over 1616060.83 frames. ], batch size: 72, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:09:56,892 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 12:10:20,608 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215687.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:10:25,967 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215695.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:10:32,545 INFO [train.py:901] (1/4) Epoch 27, batch 5550, loss[loss=0.1945, simple_loss=0.2755, pruned_loss=0.05675, over 8239.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2838, pruned_loss=0.0581, over 1617727.92 frames. ], batch size: 24, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:10:34,102 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215707.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:10:37,242 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.445e+02 2.973e+02 3.969e+02 8.778e+02, threshold=5.947e+02, percent-clipped=4.0 +2023-02-07 12:11:06,809 INFO [train.py:901] (1/4) Epoch 27, batch 5600, loss[loss=0.185, simple_loss=0.2879, pruned_loss=0.04104, over 8469.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2827, pruned_loss=0.05691, over 1619747.24 frames. ], batch size: 39, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:11:40,034 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215802.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:11:41,822 INFO [train.py:901] (1/4) Epoch 27, batch 5650, loss[loss=0.2095, simple_loss=0.2976, pruned_loss=0.06073, over 8322.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2832, pruned_loss=0.05694, over 1619962.67 frames. ], batch size: 26, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:11:46,130 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215810.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:11:47,195 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.361e+02 2.799e+02 3.308e+02 5.877e+02, threshold=5.598e+02, percent-clipped=0.0 +2023-02-07 12:11:50,151 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.0563, 1.1910, 1.1610, 0.8340, 1.2105, 0.9908, 0.1764, 1.1858], + device='cuda:1'), covar=tensor([0.0652, 0.0517, 0.0539, 0.0678, 0.0623, 0.1279, 0.1160, 0.0435], + device='cuda:1'), in_proj_covar=tensor([0.0473, 0.0410, 0.0363, 0.0459, 0.0394, 0.0552, 0.0403, 0.0439], + device='cuda:1'), out_proj_covar=tensor([1.2555e-04, 1.0640e-04, 9.4739e-05, 1.2018e-04, 1.0317e-04, 1.5401e-04, + 1.0772e-04, 1.1511e-04], device='cuda:1') +2023-02-07 12:12:04,981 WARNING [train.py:1067] (1/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 12:12:11,224 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215847.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:12:14,102 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.6835, 2.2311, 3.7040, 1.5926, 2.6632, 2.2026, 1.8667, 2.8174], + device='cuda:1'), covar=tensor([0.2023, 0.2742, 0.1062, 0.4746, 0.2189, 0.3394, 0.2527, 0.2586], + device='cuda:1'), in_proj_covar=tensor([0.0537, 0.0633, 0.0565, 0.0667, 0.0660, 0.0609, 0.0563, 0.0642], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 12:12:16,646 INFO [train.py:901] (1/4) Epoch 27, batch 5700, loss[loss=0.2076, simple_loss=0.2987, pruned_loss=0.05823, over 8248.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2835, pruned_loss=0.05744, over 1615623.74 frames. ], batch size: 24, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:12:23,687 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:12:51,905 INFO [train.py:901] (1/4) Epoch 27, batch 5750, loss[loss=0.1906, simple_loss=0.2654, pruned_loss=0.05783, over 7220.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2834, pruned_loss=0.05789, over 1615117.72 frames. ], batch size: 16, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:12:56,026 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215910.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:12:57,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.395e+02 2.899e+02 3.864e+02 7.116e+02, threshold=5.798e+02, percent-clipped=7.0 +2023-02-07 12:13:07,810 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215927.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:13:10,343 WARNING [train.py:1067] (1/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 12:13:26,393 INFO [train.py:901] (1/4) Epoch 27, batch 5800, loss[loss=0.1905, simple_loss=0.2824, pruned_loss=0.04934, over 8526.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2825, pruned_loss=0.05795, over 1610478.00 frames. ], batch size: 28, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:13:27,183 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([3.2532, 3.1884, 2.9868, 1.5617, 2.8886, 2.9479, 2.8230, 2.8346], + device='cuda:1'), covar=tensor([0.1100, 0.0773, 0.1147, 0.4509, 0.1100, 0.1347, 0.1608, 0.1054], + device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0462, 0.0441, 0.0553, 0.0441, 0.0464, 0.0441, 0.0403], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 12:13:31,934 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215963.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:13:43,178 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215980.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:13:48,230 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215988.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:13:56,129 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5923, 1.9051, 2.0359, 1.3188, 2.0801, 1.4213, 0.6358, 1.7782], + device='cuda:1'), covar=tensor([0.0765, 0.0434, 0.0369, 0.0748, 0.0529, 0.1222, 0.1139, 0.0418], + device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0411, 0.0364, 0.0460, 0.0395, 0.0554, 0.0404, 0.0440], + device='cuda:1'), out_proj_covar=tensor([1.2582e-04, 1.0678e-04, 9.4899e-05, 1.2046e-04, 1.0343e-04, 1.5464e-04, + 1.0788e-04, 1.1542e-04], device='cuda:1') +2023-02-07 12:14:01,070 INFO [train.py:901] (1/4) Epoch 27, batch 5850, loss[loss=0.1872, simple_loss=0.2588, pruned_loss=0.05785, over 7433.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2826, pruned_loss=0.0583, over 1611506.44 frames. ], batch size: 17, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:14:05,651 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.428e+02 2.871e+02 3.760e+02 7.078e+02, threshold=5.742e+02, percent-clipped=9.0 +2023-02-07 12:14:13,528 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 12:14:15,361 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216025.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:14:27,450 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216042.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:14:28,169 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7129, 1.9558, 2.0555, 1.3575, 2.1396, 1.6245, 0.5874, 1.8856], + device='cuda:1'), covar=tensor([0.0659, 0.0415, 0.0357, 0.0622, 0.0456, 0.1010, 0.1008, 0.0335], + device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0411, 0.0364, 0.0460, 0.0395, 0.0553, 0.0404, 0.0440], + device='cuda:1'), out_proj_covar=tensor([1.2583e-04, 1.0675e-04, 9.4870e-05, 1.2045e-04, 1.0353e-04, 1.5446e-04, + 1.0791e-04, 1.1531e-04], device='cuda:1') +2023-02-07 12:14:30,134 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216046.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:14:34,947 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0671, 1.8831, 2.2700, 1.9749, 2.2548, 2.1766, 1.9676, 1.2677], + device='cuda:1'), covar=tensor([0.6124, 0.5040, 0.2249, 0.4016, 0.2671, 0.3055, 0.2006, 0.5416], + device='cuda:1'), in_proj_covar=tensor([0.0963, 0.1017, 0.0831, 0.0990, 0.1025, 0.0927, 0.0769, 0.0850], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 12:14:35,103 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 12:14:36,736 INFO [train.py:901] (1/4) Epoch 27, batch 5900, loss[loss=0.1862, simple_loss=0.279, pruned_loss=0.04665, over 8330.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2822, pruned_loss=0.05777, over 1618030.15 frames. ], batch size: 25, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:14:38,990 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216058.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:14:44,146 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216066.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:14:55,474 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216083.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:15:00,973 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216091.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:15:10,165 INFO [train.py:901] (1/4) Epoch 27, batch 5950, loss[loss=0.2231, simple_loss=0.2967, pruned_loss=0.0747, over 7927.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2812, pruned_loss=0.05738, over 1615946.49 frames. ], batch size: 20, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:15:15,770 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.411e+02 2.864e+02 3.625e+02 8.908e+02, threshold=5.728e+02, percent-clipped=5.0 +2023-02-07 12:15:46,874 INFO [train.py:901] (1/4) Epoch 27, batch 6000, loss[loss=0.1984, simple_loss=0.2799, pruned_loss=0.05852, over 7975.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2809, pruned_loss=0.05724, over 1613935.01 frames. ], batch size: 21, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:15:46,874 INFO [train.py:926] (1/4) Computing validation loss +2023-02-07 12:15:59,965 INFO [train.py:935] (1/4) Epoch 27, validation: loss=0.1711, simple_loss=0.2711, pruned_loss=0.03554, over 944034.00 frames. +2023-02-07 12:15:59,966 INFO [train.py:936] (1/4) Maximum memory allocated so far is 6717MB +2023-02-07 12:16:25,737 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=216191.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:16:35,309 INFO [train.py:901] (1/4) Epoch 27, batch 6050, loss[loss=0.197, simple_loss=0.2714, pruned_loss=0.06127, over 7790.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2805, pruned_loss=0.0567, over 1615729.40 frames. ], batch size: 19, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:16:40,123 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.565e+02 3.207e+02 4.227e+02 9.285e+02, threshold=6.415e+02, percent-clipped=9.0 +2023-02-07 12:16:56,666 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216236.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:17:09,865 INFO [train.py:901] (1/4) Epoch 27, batch 6100, loss[loss=0.2155, simple_loss=0.2954, pruned_loss=0.06784, over 7047.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2814, pruned_loss=0.05709, over 1618650.12 frames. ], batch size: 71, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:17:14,012 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216261.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:17:28,454 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216281.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:17:39,621 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216298.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:17:44,974 INFO [train.py:901] (1/4) Epoch 27, batch 6150, loss[loss=0.2216, simple_loss=0.2998, pruned_loss=0.07174, over 8460.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.281, pruned_loss=0.05689, over 1617382.12 frames. ], batch size: 25, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:17:44,985 WARNING [train.py:1067] (1/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 12:17:45,829 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216306.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:17:45,852 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216306.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:17:49,771 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.311e+02 2.985e+02 4.036e+02 8.594e+02, threshold=5.970e+02, percent-clipped=2.0 +2023-02-07 12:17:57,193 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216323.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:18:18,525 INFO [train.py:901] (1/4) Epoch 27, batch 6200, loss[loss=0.1828, simple_loss=0.2631, pruned_loss=0.05123, over 8339.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2801, pruned_loss=0.05651, over 1613897.54 frames. ], batch size: 26, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:18:28,744 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 12:18:42,452 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=216390.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:18:45,892 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5347, 1.4642, 1.8323, 1.2357, 1.1801, 1.7886, 0.1934, 1.2293], + device='cuda:1'), covar=tensor([0.1558, 0.1354, 0.0419, 0.0877, 0.2305, 0.0471, 0.1950, 0.1250], + device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0205, 0.0136, 0.0222, 0.0275, 0.0145, 0.0171, 0.0199], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-02-07 12:18:53,897 INFO [train.py:901] (1/4) Epoch 27, batch 6250, loss[loss=0.2175, simple_loss=0.3006, pruned_loss=0.06721, over 8619.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2799, pruned_loss=0.05641, over 1610355.97 frames. ], batch size: 31, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:18:58,454 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.470e+02 2.901e+02 3.405e+02 7.374e+02, threshold=5.803e+02, percent-clipped=1.0 +2023-02-07 12:19:27,767 INFO [train.py:901] (1/4) Epoch 27, batch 6300, loss[loss=0.1544, simple_loss=0.2355, pruned_loss=0.03663, over 7552.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2803, pruned_loss=0.05693, over 1611609.44 frames. ], batch size: 18, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:19:54,188 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.9559, 1.5629, 1.6977, 1.4196, 1.0407, 1.5016, 1.7814, 1.4141], + device='cuda:1'), covar=tensor([0.0554, 0.1277, 0.1731, 0.1511, 0.0630, 0.1528, 0.0728, 0.0702], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0152, 0.0190, 0.0161, 0.0101, 0.0163, 0.0112, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 12:19:56,913 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.18 vs. limit=5.0 +2023-02-07 12:20:01,959 INFO [train.py:901] (1/4) Epoch 27, batch 6350, loss[loss=0.2388, simple_loss=0.3018, pruned_loss=0.0879, over 8031.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2806, pruned_loss=0.05719, over 1614479.80 frames. ], batch size: 22, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:20:02,155 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216505.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:20:07,862 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.511e+02 2.994e+02 4.018e+02 7.521e+02, threshold=5.987e+02, percent-clipped=5.0 +2023-02-07 12:20:23,599 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=3.44 vs. limit=5.0 +2023-02-07 12:20:30,861 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7145, 1.3908, 2.9402, 1.4685, 2.3344, 3.1115, 3.2605, 2.6916], + device='cuda:1'), covar=tensor([0.1135, 0.1671, 0.0332, 0.2005, 0.0782, 0.0311, 0.0708, 0.0541], + device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0325, 0.0291, 0.0318, 0.0321, 0.0276, 0.0437, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 12:20:36,668 INFO [train.py:901] (1/4) Epoch 27, batch 6400, loss[loss=0.1702, simple_loss=0.2566, pruned_loss=0.04184, over 7807.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2813, pruned_loss=0.05791, over 1612391.69 frames. ], batch size: 20, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:20:41,589 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216562.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:20:53,522 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7908, 1.8350, 1.8170, 2.3357, 1.0687, 1.5345, 1.7862, 1.8734], + device='cuda:1'), covar=tensor([0.0746, 0.0777, 0.0775, 0.0376, 0.1021, 0.1295, 0.0646, 0.0701], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0194, 0.0243, 0.0211, 0.0203, 0.0245, 0.0248, 0.0203], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 12:20:58,197 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216587.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:21:10,669 INFO [train.py:901] (1/4) Epoch 27, batch 6450, loss[loss=0.1706, simple_loss=0.2554, pruned_loss=0.0429, over 7429.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2804, pruned_loss=0.05701, over 1612954.62 frames. ], batch size: 17, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:21:13,554 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216609.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:21:16,164 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.469e+02 2.882e+02 3.609e+02 7.919e+02, threshold=5.765e+02, percent-clipped=2.0 +2023-02-07 12:21:46,032 INFO [train.py:901] (1/4) Epoch 27, batch 6500, loss[loss=0.1948, simple_loss=0.292, pruned_loss=0.04881, over 8606.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2811, pruned_loss=0.05758, over 1612468.26 frames. ], batch size: 39, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:21:58,515 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216673.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:22:19,975 INFO [train.py:901] (1/4) Epoch 27, batch 6550, loss[loss=0.1739, simple_loss=0.2697, pruned_loss=0.03904, over 8250.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2817, pruned_loss=0.05756, over 1611240.62 frames. ], batch size: 24, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:22:25,120 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.409e+02 2.544e+02 2.876e+02 3.743e+02 6.730e+02, threshold=5.752e+02, percent-clipped=5.0 +2023-02-07 12:22:32,730 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:22:48,282 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=4.14 vs. limit=5.0 +2023-02-07 12:22:52,672 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.5304, 1.7276, 2.1139, 1.7284, 1.0797, 1.8579, 2.1585, 2.0382], + device='cuda:1'), covar=tensor([0.0557, 0.1180, 0.1500, 0.1369, 0.0632, 0.1369, 0.0673, 0.0591], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0153, 0.0189, 0.0161, 0.0101, 0.0163, 0.0113, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:1') +2023-02-07 12:22:54,583 WARNING [train.py:1067] (1/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 12:22:55,725 INFO [train.py:901] (1/4) Epoch 27, batch 6600, loss[loss=0.2821, simple_loss=0.3482, pruned_loss=0.108, over 8456.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2828, pruned_loss=0.05834, over 1609169.84 frames. ], batch size: 25, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:22:59,344 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.0957, 1.9340, 2.3352, 2.0434, 2.3002, 2.0942, 2.0055, 1.5935], + device='cuda:1'), covar=tensor([0.4189, 0.3763, 0.1768, 0.3155, 0.2207, 0.2683, 0.1585, 0.4285], + device='cuda:1'), in_proj_covar=tensor([0.0966, 0.1021, 0.0832, 0.0992, 0.1029, 0.0930, 0.0773, 0.0852], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 12:22:59,994 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216761.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:23:12,955 WARNING [train.py:1067] (1/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 12:23:17,027 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216786.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:23:29,549 INFO [train.py:901] (1/4) Epoch 27, batch 6650, loss[loss=0.1721, simple_loss=0.2562, pruned_loss=0.044, over 7708.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2827, pruned_loss=0.05821, over 1610615.76 frames. ], batch size: 18, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:23:34,796 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.571e+02 3.099e+02 3.859e+02 9.745e+02, threshold=6.199e+02, percent-clipped=7.0 +2023-02-07 12:24:03,767 INFO [train.py:901] (1/4) Epoch 27, batch 6700, loss[loss=0.2244, simple_loss=0.2983, pruned_loss=0.07524, over 8316.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2828, pruned_loss=0.05823, over 1613331.47 frames. ], batch size: 26, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:24:38,627 INFO [train.py:901] (1/4) Epoch 27, batch 6750, loss[loss=0.2157, simple_loss=0.3131, pruned_loss=0.05913, over 8598.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2844, pruned_loss=0.05863, over 1615079.05 frames. ], batch size: 49, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:24:43,899 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.479e+02 3.006e+02 3.687e+02 6.813e+02, threshold=6.012e+02, percent-clipped=1.0 +2023-02-07 12:25:11,172 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=216953.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:25:12,359 INFO [train.py:901] (1/4) Epoch 27, batch 6800, loss[loss=0.2023, simple_loss=0.29, pruned_loss=0.05731, over 8101.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2857, pruned_loss=0.05919, over 1613324.35 frames. ], batch size: 23, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:25:24,107 WARNING [train.py:1067] (1/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 12:25:32,325 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216983.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:25:41,048 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=256, metric=2.99 vs. limit=5.0 +2023-02-07 12:25:47,362 INFO [train.py:901] (1/4) Epoch 27, batch 6850, loss[loss=0.1952, simple_loss=0.2874, pruned_loss=0.05146, over 8033.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2846, pruned_loss=0.05871, over 1614221.38 frames. ], batch size: 22, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:25:52,574 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.408e+02 3.097e+02 3.751e+02 9.876e+02, threshold=6.193e+02, percent-clipped=4.0 +2023-02-07 12:25:55,363 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=217017.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:26:10,981 WARNING [train.py:1067] (1/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 12:26:21,087 INFO [train.py:901] (1/4) Epoch 27, batch 6900, loss[loss=0.2212, simple_loss=0.317, pruned_loss=0.06269, over 8286.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2846, pruned_loss=0.05842, over 1612628.40 frames. ], batch size: 23, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:26:29,751 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=217067.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:26:30,521 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217068.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:26:40,219 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 12:26:56,825 INFO [train.py:901] (1/4) Epoch 27, batch 6950, loss[loss=0.1948, simple_loss=0.2878, pruned_loss=0.0509, over 8335.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2845, pruned_loss=0.05826, over 1614673.91 frames. ], batch size: 25, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:26:58,367 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([2.1888, 2.0894, 2.6545, 2.2083, 2.7985, 2.2926, 2.0947, 1.6965], + device='cuda:1'), covar=tensor([0.6143, 0.5218, 0.2123, 0.4436, 0.2661, 0.3287, 0.2050, 0.5647], + device='cuda:1'), in_proj_covar=tensor([0.0966, 0.1020, 0.0833, 0.0992, 0.1030, 0.0928, 0.0772, 0.0850], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-02-07 12:27:02,039 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.303e+02 2.670e+02 3.410e+02 6.861e+02, threshold=5.340e+02, percent-clipped=1.0 +2023-02-07 12:27:15,657 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217132.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:27:19,468 WARNING [train.py:1067] (1/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 12:27:30,825 INFO [train.py:901] (1/4) Epoch 27, batch 7000, loss[loss=0.25, simple_loss=0.327, pruned_loss=0.08646, over 8500.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2839, pruned_loss=0.0581, over 1610613.84 frames. ], batch size: 49, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:27:31,047 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8897, 2.4920, 4.1384, 1.7154, 3.0284, 2.4606, 1.9739, 3.0800], + device='cuda:1'), covar=tensor([0.1849, 0.2499, 0.0821, 0.4620, 0.1795, 0.3109, 0.2327, 0.2287], + device='cuda:1'), in_proj_covar=tensor([0.0536, 0.0630, 0.0561, 0.0664, 0.0655, 0.0607, 0.0561, 0.0641], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 12:27:49,479 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217182.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:27:54,192 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([4.1524, 1.3677, 4.3532, 1.6623, 3.8566, 3.6687, 3.9922, 3.8737], + device='cuda:1'), covar=tensor([0.0695, 0.5025, 0.0552, 0.4317, 0.1148, 0.0954, 0.0610, 0.0653], + device='cuda:1'), in_proj_covar=tensor([0.0673, 0.0661, 0.0730, 0.0653, 0.0743, 0.0627, 0.0632, 0.0713], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 12:28:05,136 INFO [train.py:901] (1/4) Epoch 27, batch 7050, loss[loss=0.2377, simple_loss=0.3139, pruned_loss=0.08071, over 8314.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2849, pruned_loss=0.05915, over 1608087.32 frames. ], batch size: 25, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:28:11,286 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.356e+02 3.046e+02 3.591e+02 8.726e+02, threshold=6.092e+02, percent-clipped=6.0 +2023-02-07 12:28:40,066 INFO [train.py:901] (1/4) Epoch 27, batch 7100, loss[loss=0.2033, simple_loss=0.2917, pruned_loss=0.0575, over 8326.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2844, pruned_loss=0.05891, over 1610365.93 frames. ], batch size: 25, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:29:10,625 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.5400, 2.0517, 2.8696, 1.5120, 2.1904, 1.9734, 1.7536, 2.1837], + device='cuda:1'), covar=tensor([0.2038, 0.2550, 0.0958, 0.4685, 0.1963, 0.3381, 0.2395, 0.2463], + device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0633, 0.0564, 0.0667, 0.0657, 0.0610, 0.0564, 0.0643], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 12:29:14,484 INFO [train.py:901] (1/4) Epoch 27, batch 7150, loss[loss=0.1984, simple_loss=0.2849, pruned_loss=0.05595, over 8583.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2842, pruned_loss=0.05841, over 1615343.98 frames. ], batch size: 39, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:29:19,666 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 2.444e+02 3.123e+02 4.113e+02 1.134e+03, threshold=6.246e+02, percent-clipped=7.0 +2023-02-07 12:29:27,888 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217324.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:29:29,712 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=217327.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:29:45,942 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217349.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:29:49,821 INFO [train.py:901] (1/4) Epoch 27, batch 7200, loss[loss=0.1877, simple_loss=0.2696, pruned_loss=0.05287, over 7792.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2837, pruned_loss=0.05766, over 1613683.76 frames. ], batch size: 19, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:30:11,976 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217388.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:30:23,029 INFO [train.py:901] (1/4) Epoch 27, batch 7250, loss[loss=0.2038, simple_loss=0.2941, pruned_loss=0.05672, over 8359.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2842, pruned_loss=0.05799, over 1612540.99 frames. ], batch size: 24, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:30:23,857 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217406.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:30:28,396 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.296e+02 2.784e+02 3.610e+02 7.832e+02, threshold=5.568e+02, percent-clipped=2.0 +2023-02-07 12:30:28,621 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217413.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:30:45,935 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217438.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:30:49,265 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217442.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:30:58,441 INFO [train.py:901] (1/4) Epoch 27, batch 7300, loss[loss=0.1593, simple_loss=0.2495, pruned_loss=0.03459, over 8131.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.284, pruned_loss=0.05795, over 1616130.32 frames. ], batch size: 22, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:31:04,048 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217463.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:31:33,102 INFO [train.py:901] (1/4) Epoch 27, batch 7350, loss[loss=0.1875, simple_loss=0.2804, pruned_loss=0.04728, over 8238.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2834, pruned_loss=0.05772, over 1611738.26 frames. ], batch size: 22, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:31:36,625 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217510.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:31:38,501 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.663e+02 2.322e+02 2.888e+02 3.768e+02 6.651e+02, threshold=5.777e+02, percent-clipped=4.0 +2023-02-07 12:31:59,847 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 12:32:07,063 INFO [train.py:901] (1/4) Epoch 27, batch 7400, loss[loss=0.209, simple_loss=0.2926, pruned_loss=0.06265, over 8205.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2816, pruned_loss=0.05699, over 1611061.37 frames. ], batch size: 23, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:32:19,152 WARNING [train.py:1067] (1/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 12:32:42,460 INFO [train.py:901] (1/4) Epoch 27, batch 7450, loss[loss=0.2021, simple_loss=0.2787, pruned_loss=0.06269, over 7793.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.283, pruned_loss=0.05784, over 1612209.31 frames. ], batch size: 19, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:32:43,319 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.8659, 1.3801, 3.5494, 1.6748, 2.5177, 3.8230, 3.9807, 3.2841], + device='cuda:1'), covar=tensor([0.1279, 0.1942, 0.0270, 0.1929, 0.0905, 0.0221, 0.0576, 0.0557], + device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0322, 0.0291, 0.0318, 0.0320, 0.0276, 0.0437, 0.0306], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:1') +2023-02-07 12:32:47,778 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.478e+02 3.262e+02 4.062e+02 8.102e+02, threshold=6.523e+02, percent-clipped=5.0 +2023-02-07 12:32:58,342 WARNING [train.py:1067] (1/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 12:33:05,996 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 12:33:16,123 INFO [train.py:901] (1/4) Epoch 27, batch 7500, loss[loss=0.2205, simple_loss=0.3045, pruned_loss=0.06828, over 8613.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2832, pruned_loss=0.05753, over 1618611.61 frames. ], batch size: 31, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:33:34,969 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217682.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:33:46,549 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217698.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:33:51,450 INFO [train.py:901] (1/4) Epoch 27, batch 7550, loss[loss=0.2415, simple_loss=0.3219, pruned_loss=0.08055, over 8660.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2829, pruned_loss=0.05758, over 1618002.52 frames. ], batch size: 34, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:33:56,752 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.627e+02 2.428e+02 3.024e+02 3.911e+02 8.560e+02, threshold=6.047e+02, percent-clipped=1.0 +2023-02-07 12:34:03,669 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:34:21,849 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=217750.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:34:25,196 INFO [train.py:901] (1/4) Epoch 27, batch 7600, loss[loss=0.199, simple_loss=0.2839, pruned_loss=0.05705, over 8296.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2827, pruned_loss=0.05732, over 1614160.90 frames. ], batch size: 23, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:35:01,505 INFO [train.py:901] (1/4) Epoch 27, batch 7650, loss[loss=0.1695, simple_loss=0.259, pruned_loss=0.04002, over 7969.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2829, pruned_loss=0.05739, over 1613211.29 frames. ], batch size: 21, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:35:06,795 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.541e+02 2.896e+02 3.920e+02 6.720e+02, threshold=5.793e+02, percent-clipped=4.0 +2023-02-07 12:35:35,073 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=217854.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:35:35,687 INFO [train.py:901] (1/4) Epoch 27, batch 7700, loss[loss=0.1915, simple_loss=0.2837, pruned_loss=0.04971, over 8496.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2831, pruned_loss=0.05756, over 1616492.28 frames. ], batch size: 26, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:35:42,369 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:35:54,057 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 +2023-02-07 12:36:05,167 WARNING [train.py:1067] (1/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 12:36:10,561 INFO [train.py:901] (1/4) Epoch 27, batch 7750, loss[loss=0.1848, simple_loss=0.2767, pruned_loss=0.04645, over 8328.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2831, pruned_loss=0.0578, over 1617105.15 frames. ], batch size: 25, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:36:15,958 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.515e+02 3.033e+02 3.634e+02 8.452e+02, threshold=6.066e+02, percent-clipped=4.0 +2023-02-07 12:36:18,821 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217916.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:36:45,567 INFO [train.py:901] (1/4) Epoch 27, batch 7800, loss[loss=0.2392, simple_loss=0.3124, pruned_loss=0.08299, over 8511.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.282, pruned_loss=0.05681, over 1617914.05 frames. ], batch size: 26, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:36:55,073 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217969.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:37:08,175 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.7950, 2.6743, 1.8659, 2.3552, 2.2131, 1.7171, 2.1957, 2.4174], + device='cuda:1'), covar=tensor([0.1527, 0.0383, 0.1203, 0.0720, 0.0812, 0.1517, 0.1133, 0.0999], + device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0241, 0.0341, 0.0312, 0.0303, 0.0346, 0.0349, 0.0322], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-02-07 12:37:17,814 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([1.4000, 1.4413, 1.3475, 1.8413, 0.7002, 1.2431, 1.3447, 1.4530], + device='cuda:1'), covar=tensor([0.0857, 0.0790, 0.0994, 0.0513, 0.1114, 0.1364, 0.0716, 0.0705], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0193, 0.0245, 0.0212, 0.0202, 0.0245, 0.0249, 0.0203], + device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:1') +2023-02-07 12:37:19,657 INFO [train.py:901] (1/4) Epoch 27, batch 7850, loss[loss=0.2168, simple_loss=0.2976, pruned_loss=0.06799, over 7525.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2826, pruned_loss=0.05742, over 1616971.05 frames. ], batch size: 71, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:37:24,964 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.749e+02 2.277e+02 2.828e+02 3.912e+02 8.712e+02, threshold=5.655e+02, percent-clipped=7.0 +2023-02-07 12:37:33,506 INFO [zipformer.py:1185] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=218026.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:37:52,845 INFO [train.py:901] (1/4) Epoch 27, batch 7900, loss[loss=0.1553, simple_loss=0.243, pruned_loss=0.03383, over 7926.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2813, pruned_loss=0.05701, over 1612983.99 frames. ], batch size: 20, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:38:26,204 INFO [train.py:901] (1/4) Epoch 27, batch 7950, loss[loss=0.2137, simple_loss=0.3027, pruned_loss=0.0623, over 8334.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2821, pruned_loss=0.0575, over 1612794.86 frames. ], batch size: 26, lr: 2.76e-03, grad_scale: 8.0 +2023-02-07 12:38:31,700 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.553e+02 3.230e+02 4.059e+02 8.354e+02, threshold=6.459e+02, percent-clipped=5.0 +2023-02-07 12:38:35,248 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:38:37,406 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218121.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:38:50,509 INFO [zipformer.py:1185] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218141.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:38:53,612 INFO [zipformer.py:1185] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218146.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:38:58,211 INFO [zipformer.py:2431] (1/4) attn_weights_entropy = tensor([5.8262, 6.0254, 5.1958, 2.1805, 5.2848, 5.6439, 5.4834, 5.4146], + device='cuda:1'), covar=tensor([0.0486, 0.0382, 0.1042, 0.4830, 0.0759, 0.0938, 0.1037, 0.0566], + device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0462, 0.0447, 0.0557, 0.0442, 0.0464, 0.0440, 0.0407], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-02-07 12:38:59,474 INFO [train.py:901] (1/4) Epoch 27, batch 8000, loss[loss=0.2148, simple_loss=0.2972, pruned_loss=0.06616, over 8659.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2821, pruned_loss=0.0573, over 1615890.20 frames. ], batch size: 34, lr: 2.76e-03, grad_scale: 8.0 +2023-02-07 12:39:29,224 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218200.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:39:32,352 INFO [train.py:901] (1/4) Epoch 27, batch 8050, loss[loss=0.173, simple_loss=0.2483, pruned_loss=0.04884, over 7428.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2811, pruned_loss=0.05688, over 1611211.90 frames. ], batch size: 17, lr: 2.76e-03, grad_scale: 8.0 +2023-02-07 12:39:38,065 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.285e+02 2.948e+02 3.498e+02 7.136e+02, threshold=5.897e+02, percent-clipped=2.0 +2023-02-07 12:39:46,230 INFO [zipformer.py:1185] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218225.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:39:48,249 INFO [zipformer.py:1185] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218228.0, num_to_drop=0, layers_to_drop=set()